# Shadow AI Detection Tools: every page as markdown Ranked comparison of 16 shadow AI detection tools, scored on discovery, vantage points, speed, actions, context, transparency and proof. Updated 16 Nov 2026. --- # Shadow AI detection tools, ranked and compared URL: https://shadowaidetectiontools.com/ Updated: 2026-11-16 Harmonic Security leads the ranking of 16 shadow AI detection tools with 73.5 points out of 100, scored on seven fixed criteria from vendor pages, documentation and third-party reports read on 16 Nov 2026. Bay's position is joint 11th, at 49.5 points. Pick a situation below to see a different order. ## The short answer Harmonic Security leads the site ranking at 73.5 of 100. Bay is joint 11th at 49.5. Harmonic Security is ahead mainly on where it looks (7 against 4, worth 6.0 points at weight 20). ## What is on your machines Each row is a kind of AI tool. The table lists the vendors that document finding it, the ones that document it in part, and the ones with nothing found on 2026-11-16. With scripts on, tick the kinds your staff use and the result below lists the tools that find all of them. **Kinds of AI tool and which vendors find them** | Kind of AI tool | Documented | Partial | None found | |---|---|---|---| | GenAI web apps | [Aim Security](/vendors/aim-security/#fact-aim-security-10), [Aurascape](/vendors/aurascape/#fact-aurascape-06), [Bay](/vendors/bay/#fact-bay-12), [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-01), [Grip Security](/vendors/grip-security/#fact-grip-security-05), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-10), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-06), [Microsoft](/vendors/microsoft/#fact-microsoft-01), [Netskope](/vendors/netskope/#fact-netskope-04), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-01), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-05), [Push Security](/vendors/push-security/#fact-push-security-03), [Reco](/vendors/reco/#fact-reco-02), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-08), [WitnessAI](/vendors/witnessai/#fact-witnessai-05), [Zscaler](/vendors/zscaler/#fact-zscaler-08) | none | none | | Desktop AI apps | [Aim Security](/vendors/aim-security/#fact-aim-security-12), [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-04), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-12), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-06), [Microsoft](/vendors/microsoft/#fact-microsoft-11), [WitnessAI](/vendors/witnessai/#fact-witnessai-06), [Zscaler](/vendors/zscaler/#fact-zscaler-11) | [Aurascape](/vendors/aurascape/#fact-aurascape-15), [Bay](/vendors/bay/#fact-bay-14), [Netskope](/vendors/netskope/#fact-netskope-05), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-06), [Push Security](/vendors/push-security/#fact-push-security-08), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-15) | [Grip Security](/vendors/grip-security/#fact-grip-security-09), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-04), [Reco](/vendors/reco/#fact-reco-09) | | Coding agents | [Aim Security](/vendors/aim-security/#fact-aim-security-11), [Aurascape](/vendors/aurascape/#fact-aurascape-09), [Bay](/vendors/bay/#fact-bay-14), [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-01), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-12), [Microsoft](/vendors/microsoft/#fact-microsoft-11), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-09), [WitnessAI](/vendors/witnessai/#fact-witnessai-07) | [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-07), [Netskope](/vendors/netskope/#fact-netskope-04), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-38), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-05), [Push Security](/vendors/push-security/#fact-push-security-03), [Zscaler](/vendors/zscaler/#fact-zscaler-15) | [Grip Security](/vendors/grip-security/#fact-grip-security-09), [Reco](/vendors/reco/#fact-reco-09) | | Browser extensions | [Bay](/vendors/bay/#fact-bay-12), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-08), [Netskope](/vendors/netskope/#fact-netskope-05), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-03), [Push Security](/vendors/push-security/#fact-push-security-04), [Reco](/vendors/reco/#fact-reco-03) | [Aim Security](/vendors/aim-security/#fact-aim-security-34), [Aurascape](/vendors/aurascape/#fact-aurascape-09), [Grip Security](/vendors/grip-security/#fact-grip-security-05), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-15), [Microsoft](/vendors/microsoft/#fact-microsoft-14), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-08), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-11), [WitnessAI](/vendors/witnessai/#fact-witnessai-28), [Zscaler](/vendors/zscaler/#fact-zscaler-06) | [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-07) | | MCP servers | [Aim Security](/vendors/aim-security/#fact-aim-security-11), [Aurascape](/vendors/aurascape/#fact-aurascape-08), [Bay](/vendors/bay/#fact-bay-14), [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-02), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-14), [Microsoft](/vendors/microsoft/#fact-microsoft-15), [Netskope](/vendors/netskope/#fact-netskope-04), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-10), [WitnessAI](/vendors/witnessai/#fact-witnessai-08) | [Nudge Security](/vendors/nudge-security/#fact-nudge-security-05), [Push Security](/vendors/push-security/#fact-push-security-07), [Zscaler](/vendors/zscaler/#fact-zscaler-13) | [Grip Security](/vendors/grip-security/#fact-grip-security-09), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-11), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-11), [Reco](/vendors/reco/#fact-reco-09) | | Embedded SaaS AI | [Aurascape](/vendors/aurascape/#fact-aurascape-10), [Bay](/vendors/bay/#fact-bay-11), [Grip Security](/vendors/grip-security/#fact-grip-security-06), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-10), [Netskope](/vendors/netskope/#fact-netskope-08), [Reco](/vendors/reco/#fact-reco-04), [Zscaler](/vendors/zscaler/#fact-zscaler-10) | [Aim Security](/vendors/aim-security/#fact-aim-security-14), [Microsoft](/vendors/microsoft/#fact-microsoft-01), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-06), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-11), [WitnessAI](/vendors/witnessai/#fact-witnessai-06) | [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-07), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-11), [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-11), [Push Security](/vendors/push-security/) | | OAuth AI apps | [Grip Security](/vendors/grip-security/#fact-grip-security-07), [Microsoft](/vendors/microsoft/#fact-microsoft-06), [Nudge Security](/vendors/nudge-security/#fact-nudge-security-01), [Push Security](/vendors/push-security/#fact-push-security-05), [Reco](/vendors/reco/#fact-reco-03) | [Palo Alto Networks](/vendors/palo-alto-networks/#fact-palo-alto-networks-07) | [Aim Security](/vendors/aim-security/#fact-aim-security-15), [Aurascape](/vendors/aurascape/#fact-aurascape-11), [Bay](/vendors/bay/#fact-bay-32), [Cyberhaven](/vendors/cyberhaven/#fact-cyberhaven-07), [Harmonic Security](/vendors/harmonic-security/#fact-harmonic-security-16), [LayerX (Akamai Workforce Protector)](/vendors/layerx/#fact-layerx-11), [Netskope](/vendors/netskope/#fact-netskope-07), [SentinelOne (Prompt Security)](/vendors/sentinelone/#fact-sentinelone-12), [WitnessAI](/vendors/witnessai/#fact-witnessai-10), [Zscaler](/vendors/zscaler/#fact-zscaler-26) | Tick one or more kinds above. ## The ranking [The full ranking](/ranking/) has the same table with re-rank sliders. Each score links to the evidence on the vendor's page. **Site ranking: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 25) | Where it looks (weight 20) | Time to a first inventory (weight 15) | What it can do about a finding (weight 15) | Risk context and reporting (weight 10) | Transparency (weight 10) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 73.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | 2nd | [Microsoft](/vendors/microsoft/) | 66.5 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | 3rd | [Netskope](/vendors/netskope/) | 65.5 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 4th | [Nudge Security](/vendors/nudge-security/) | 64.5 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 5th | [Grip Security](/vendors/grip-security/) | 58.5 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | 6th | [Zscaler](/vendors/zscaler/) | 56.5 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | 7th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 55.5 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 8th | [Cyberhaven](/vendors/cyberhaven/) | 54.5 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | joint 9th | [Aurascape](/vendors/aurascape/) | 53.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | joint 9th | [Push Security](/vendors/push-security/) | 53.5 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | joint 11th | [Bay](/vendors/bay/) | 49.5 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | joint 11th | [Reco](/vendors/reco/) | 49.5 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 48.0 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 14th | [Aim Security](/vendors/aim-security/) | 47.5 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 15th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 45.5 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 42.0 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## Who should pick what With no particular situation, the published ranking applies and Harmonic Security comes first. For a specific situation the weights change, and so can the order: - **[A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/)**: pick Harmonic Security (76.0). Bay is joint 13th at 48.5. An auditor asks for a complete inventory with risk ratings and an evidence trail, so breadth of discovery and the risk report weigh most. - **[A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/)**: pick Harmonic Security (73.5). Bay is joint 6th at 55.0. A hard deadline rewards the widest discovery that is running fastest; acting on findings can follow. - **[An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/)**: pick Harmonic Security (72.5). Bay is joint 10th at 52.0. Developer tools live on the endpoint and in the IDE, out of sight of SaaS logs, so what is discovered and where the tool looks matter most. - **[A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/)**: pick Harmonic Security (73.5). Bay is 12th at 50.0. Embedded AI and OAuth grants show up in identity and SaaS records, so reading those vantage points weighs most. - **[A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/)**: pick Harmonic Security (75.5). Bay is 11th at 53.0. Guiding people in the moment is the job, so what the tool can do when it finds an unapproved app weighs most. ## What would change this ranking **Top pair.** Harmonic Security falling from 7 to 4 on what it discovers (weight 25, so each point moves a total by 2.5) would put Microsoft above Harmonic Security; the gap is 7.0 points. **Bay's next place.** Bay rising from 6 to 8 on what it discovers (weight 25, so each point moves a total by 2.5) would put Bay above Push Security; the gap is 4.0 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## What are the best shadow AI detection tools? Harmonic Security is first in the published ranking with 73.5 points out of 100, and Bay's position is joint 11th with 49.5. The ranking covers 16 tools that find the AI web apps, desktop apps, coding agents, browser extensions, IDE plugins, MCP servers and embedded SaaS features that staff use without approval. Every tool is scored 0 to 10 on seven criteria whose weights were committed before any score: what it discovers (25), where it looks (20), time to a first inventory (15), what it can do about a finding (15), risk context and reporting (10), transparency (10) and proof (5). A vendor's own claim earns at most 7 until a second source confirms it. The right choice depends on the situation. A regulated company, a CISO with a 30-day board deadline, an engineering organization, a SaaS-heavy company and a team that wants to coach rather than block each weigh the criteria differently, and each has its own ranked page. ## How are shadow AI detection tools ranked here? Each tool is scored from a fixed evidence file gathered on 16 Nov 2026, and the total is the weighted sum of the seven scores. Facts come from vendor pages, documentation, pricing pages, release notes and at least two third-party hosts per vendor. A scorer worked from the evidence with vendor names replaced by codes, and a second reviewer checked the scores blind. The method page lists the weights, the anchors for each score and the rules, and the re-rank sliders let a reader change the weights to match their own priorities. ## Which shadow AI detection tool finds the most kinds of AI? No vendor documents all seven kinds as found. Harmonic Security, Microsoft and Bay each show five of seven kinds as documented, though Bay's coverage rests on vendor claims alone and Microsoft's five include apps granted OAuth access but not embedded SaaS AI. The seven kinds are AI web apps, desktop AI apps, coding agents, browser extensions, MCP servers, AI embedded in approved SaaS and AI apps granted OAuth access. Tick the kinds present on your machines in the tool above to see which products document each one, and read the vendor profile for the source behind every cell. ## Latest news, guides and data - [Nudge Security adds OpenAI Agents connected app](/news/nudge-security-adds-openai-agents-connected-app/) (2026-11-16) - [Nudge Security adds regional data checks for AI chats](/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/) (2026-11-09) - [Cyberhaven launches Agentic Data Security module](/news/cyberhaven-launches-agentic-data-security/) (2026-11-05) - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [Vantage Point Matrix](/vantage-point-matrix/), with [CSV and JSON downloads](/datasets/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/), with [CSV and JSON downloads](/datasets/finding-to-action-ladder/) ## Related pages - [Full ranking with re-rank sliders](/ranking/) - [How the scores and weights work](/method/) - [Vendor profiles with evidence](/vendors/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) - [Guides](/guides/) - [Frequently asked questions](/faq/) --- # Shadow AI detection tools ranking URL: https://shadowaidetectiontools.com/ranking/ Updated: 2026-11-16 Harmonic Security ranks first among 16 shadow AI detection tools with 73.5 of 100 points on the weights fixed on 16 Nov 2026. Bay is at joint 11th with 49.5. Totals are weighted sums of seven 0 to 10 scores, and joint ranks are printed where totals are equal. ## The short answer Harmonic Security leads the site ranking at 73.5 of 100. Bay is joint 11th at 49.5. Harmonic Security is ahead mainly on where it looks (7 against 4, worth 6.0 points at weight 20). **Site ranking: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 25) | Where it looks (weight 20) | Time to a first inventory (weight 15) | What it can do about a finding (weight 15) | Risk context and reporting (weight 10) | Transparency (weight 10) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 73.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | 2nd | [Microsoft](/vendors/microsoft/) | 66.5 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | 3rd | [Netskope](/vendors/netskope/) | 65.5 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 4th | [Nudge Security](/vendors/nudge-security/) | 64.5 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 5th | [Grip Security](/vendors/grip-security/) | 58.5 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | 6th | [Zscaler](/vendors/zscaler/) | 56.5 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | 7th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 55.5 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 8th | [Cyberhaven](/vendors/cyberhaven/) | 54.5 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | joint 9th | [Aurascape](/vendors/aurascape/) | 53.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | joint 9th | [Push Security](/vendors/push-security/) | 53.5 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | joint 11th | [Bay](/vendors/bay/) | 49.5 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | joint 11th | [Reco](/vendors/reco/) | 49.5 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 48.0 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 14th | [Aim Security](/vendors/aim-security/) | 47.5 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 15th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 45.5 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 42.0 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## Who should pick what With no particular situation, the published ranking applies and Harmonic Security comes first. For a specific situation the weights change, and so can the order: - **[A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/)**: pick Harmonic Security (76.0). Bay is joint 13th at 48.5. An auditor asks for a complete inventory with risk ratings and an evidence trail, so breadth of discovery and the risk report weigh most. - **[A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/)**: pick Harmonic Security (73.5). Bay is joint 6th at 55.0. A hard deadline rewards the widest discovery that is running fastest; acting on findings can follow. - **[An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/)**: pick Harmonic Security (72.5). Bay is joint 10th at 52.0. Developer tools live on the endpoint and in the IDE, out of sight of SaaS logs, so what is discovered and where the tool looks matter most. - **[A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/)**: pick Harmonic Security (73.5). Bay is 12th at 50.0. Embedded AI and OAuth grants show up in identity and SaaS records, so reading those vantage points weighs most. - **[A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/)**: pick Harmonic Security (75.5). Bay is 11th at 53.0. Guiding people in the moment is the job, so what the tool can do when it finds an unapproved app weighs most. ## What would change this ranking **Top pair.** Harmonic Security falling from 7 to 4 on what it discovers (weight 25, so each point moves a total by 2.5) would put Microsoft above Harmonic Security; the gap is 7.0 points. **Bay's next place.** Bay rising from 6 to 8 on what it discovers (weight 25, so each point moves a total by 2.5) would put Bay above Push Security; the gap is 4.0 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which shadow AI detection tool ranks first? Harmonic Security ranks first with 73.5 points out of 100, and the table above lists all 16 tools with their totals. Bay is at joint 11th with 49.5 points. A total is the weighted sum of seven scores divided by 10. Where two totals are equal the rank is joint and the tools are listed alphabetically for display only. No score changes unless a sourced fact changes, and the change is logged with the fact. ## How do shadow AI discovery tools differ by where they look? They differ mostly by vantage point: the endpoint, the browser, network or proxy traffic, the identity provider and SaaS APIs, or email and billing records. Each vantage point sees something the others miss, so the choice follows where staff actually use AI. Endpoint and browser agents (Harmonic Security, Cyberhaven, SentinelOne, Bay, LayerX) see local apps, coding agents and MCP servers. Network and proxy tools (Zscaler, Palo Alto Networks, Aurascape, WitnessAI, Netskope) see traffic without touching devices. Identity and SaaS API tools (Grip Security, Reco, Nudge Security) see OAuth grants and embedded features. Microsoft documents endpoint, browser, network and SaaS API feeds across separate products. The Vantage Point Matrix shows every cell. ## How fast does each tool produce a first inventory? Few vendors publish a figure. Bay says under 5 minutes via one MDM or EDR script, Grip Security says 10 minutes of setup, Harmonic says first insight in about 30 minutes, and Nudge Security says most organizations deploy in under an hour. Reco publishes 24 to 48 hours to full deployment. Push Security says per-user self-enrollment takes about a minute. For Cyberhaven, Netskope, Zscaler, Palo Alto Networks, Aurascape, WitnessAI, Aim Security, LayerX and Microsoft the figure is none found on 16 Nov 2026. All published figures are vendor claims, and the speed score reflects that. ## Related pages - [Method: weights, anchors and rules](/method/) - [Rankings by situation](/best-for/) - [Vendor profiles](/vendors/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) - [Guides for buyers](/guides/) - [Glossary](/glossary/) --- # How shadow AI detection tools are scored URL: https://shadowaidetectiontools.com/method/ Updated: 2026-11-16 One research plan was run for all 16 vendors on 16 Nov 2026, producing 1,469 facts across 32 vendor files, each quote checked against its page. A scorer worked blind from those files, a second blind reviewer checked every score, and the weights were committed before any evidence or score existed. ## The weights The weights were fixed on 2026-11-16, before any score existed, and have not changed since. **Criteria and weights (out of 100)** | Criterion | What it measures | Weight | |---|---|---| | What it discovers | The kinds of unapproved AI it finds, as documented: AI web apps, desktop AI apps, coding assistants and command-line agents, browser extensions, IDE plugins, MCP servers, AI features embedded in SaaS, and AI apps granted OAuth access | 25 | | Where it looks | The vantage points it reads from, as documented: the endpoint, the browser, network or proxy traffic, the identity provider, SaaS APIs and logs, email or billing records; more vantage points and named integrations score higher | 20 | | Time to a first inventory | How quickly a first AI inventory appears after setup, as published, and how light the setup is (agentless, API connection, extension or agent), with operating systems and browsers covered | 15 | | What it can do about a finding | Actions on a discovered tool, as documented: block, warn or coach the user in the moment, require approval, revoke OAuth grants, redact or stop data sent to the tool, route to a ticket | 15 | | Risk context and reporting | A risk rating per AI tool found (data use, training on inputs, vendor posture), who uses it and how much, and an exportable inventory or audit report | 10 | | Transparency | Public documentation readable without sign-in, a published price, published research | 10 | | Proof | Named customers, funding, independent analyst coverage | 5 | ## Weights by situation **Weights used for each situation view** | Situation | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---|---| | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 25 | 20 | 5 | 15 | 25 | 5 | 5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 30 | 20 | 30 | 5 | 5 | 5 | 5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 35 | 25 | 10 | 15 | 5 | 5 | 5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 25 | 30 | 10 | 15 | 10 | 5 | 5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 20 | 15 | 10 | 35 | 10 | 5 | 5 | ## Score anchors - **0**: none found on the read date - **1-3**: mentioned, partial, or limited to one surface, one layer or one action type - **4-6**: documented and working, with gaps the documentation shows - **7**: the most a vendor claim or vendor documentation alone can earn: detailed, specific, covering the whole criterion - **8-10**: detailed and confirmed by a second source (third-party write-up, or observed directly in public documentation, code, a marketplace listing or a release record) ## Rules - Every vendor is scored 0 to 10 on every criterion, whole numbers, from the evidence file only, by a scorer that works from coded packets with every vendor name removed. - A vendor claim scores at most 7 until a second source confirms it (kind third-party or observed). - None found scores 0 and is written "none found on the read date". - Breadth facts (module counts, integration counts, headcount, funding) inform Proof only, which weighs 5, so they carry at most 10 percent of any ranking. - The total is the weighted sum of the scores divided by 10 (0 to 100 scale, published to one decimal). - Equal totals print as a joint rank (for example joint 2nd); inside a joint rank vendors are listed alphabetically for display only. - Once the first score exists, no weight and no criterion changes. A score changes only when a sourced fact changes, logged with the fact. ## How were the facts gathered? One research plan was applied to every vendor on 16 Nov 2026: the vendor's own pages, documentation, pricing, release notes and at least two third-party hosts, such as news outlets and analyst or marketplace listings. Each fact records its claim, a verbatim quote, a source URL, the read date and a kind (vendor, third-party, observed or none found). The run produced 1,469 facts across 32 vendor files (758 of them in this site's 16 files), and each quote was checked against its source. Where nothing was found, the file says none found on 16 Nov 2026 instead of leaving a gap, so an absence is visible and sourced. ## How are shadow AI detection tools scored? Each vendor receives a whole-number score from 0 to 10 on seven criteria: what it discovers, where it looks, time to a first inventory, what it can do about a finding, risk context and reporting, transparency, and proof. The total is the weighted sum of the scores divided by 10, published to one decimal on a 0 to 100 scale. Scoring was blind. One scorer worked only from the evidence files with vendor names replaced by codes, then one blind reviewer checked each score against the same files. Every score carries the fact IDs it relies on and a one-sentence reason, both printed on the vendor page. ## Why were the weights fixed before any score? The weights and criteria were committed on 16 Nov 2026 before any vendor evidence was gathered and before any score existed, so the scale could not be adjusted to favor a result. They are: discovers 25, where 20, time to a first inventory 15, acts 15, context 10, transparency 10, proof 5. Once the first score existed, no weight and no criterion changed. A score changes only when a sourced fact changes, and the change is logged with that fact in the changelog. Five situations reweight the same scores for different buyers, and their weights were fixed in the same commit. ## What counts as evidence? Evidence is a sourced fact in a vendor's file. A vendor claim scores at most 7 on a criterion until a second source, either third-party coverage or something observed directly in public documentation, a marketplace listing or a release record, confirms it. A score of 8 to 10 needs that confirmation. None found scores 0 and is written as none found on the read date. Breadth facts such as module counts, integration counts, headcount and funding inform Proof only, which weighs 5, so they carry at most 10 percent of any ranking. The anchors table below lists the meaning of each score band. ## How do I request a correction? Email editor@shadowaidetectiontools.com with the page, the sentence or number, and a source URL that shows the fact. A vendor may also send a primary source that supports a higher score. A corrected fact is re-read, the score is revised only if the fact changes, and the change is recorded with the date in the changelog and on the corrections page. ## Changelog - 2026-11-16: weights and criteria fixed. - 2026-11-16: evidence read for every vendor by one research plan. - 2026-11-16: scores published (blind: one scorer, then one blind reviewer). - 2026-11-16: this site built from the committed scores. ## Related pages - [The ranking these weights produce](/ranking/) - [Vendor profiles with evidence per score](/vendors/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [Glossary](/glossary/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Shadow AI detection vendors URL: https://shadowaidetectiontools.com/vendors/ Updated: 2026-11-16 Sixteen shadow AI detection vendors are profiled: Aim Security, Aurascape, Bay, Cyberhaven, Grip Security, Harmonic Security, LayerX, Microsoft, Netskope, Nudge Security, Palo Alto Networks, Push Security, Reco, SentinelOne, WitnessAI and Zscaler. Each profile lists scores, reasons and facts read on 16 Nov 2026. ## All 16 vendors **Vendors in the ranking, with their rank and total** | Vendor | Rank | Total (of 100) | Product assessed | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | 14th | 47.5 | Cato AI Security for End Users with Cato AI Endpoint (shadow AI and agent discovery) | | [Aurascape](/vendors/aurascape/) | joint 9th | 53.5 | Aurascape AI Security Platform, Discover and Monitor AI and Control AI Access and Usage (AI Proxy based discovery and risk scoring) | | [Bay](/vendors/bay/) | joint 11th | 49.5 | Bay Shadow AI Discovery (AI Asset Discovery) | | [Cyberhaven](/vendors/cyberhaven/) | 8th | 54.5 | Cyberhaven Flow: Agentic Data Security (AI security: shadow AI discovery, AI Risk IQ scoring, AI data flow control), with the endpoint agent and browser extension | | [Grip Security](/vendors/grip-security/) | 5th | 58.5 | Grip AI + SaaS Security Control Plane: Shadow AI Discovery, free Shadow AI Analysis, Grip Extend | | [Harmonic Security](/vendors/harmonic-security/) | 1st | 73.5 | Harmonic Explore (discovery and risk scoring) and Harmonic Guide (inline coaching), with the Endpoint/Protect Agent and browser extension | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 15th | 45.5 | Akamai Workforce Protector (formerly LayerX): browser extension, management console and cloud intelligence | | [Microsoft](/vendors/microsoft/) | 2nd | 66.5 | Microsoft Defender for Cloud Apps (cloud discovery, app governance), Defender for Endpoint local AI agent discovery, Microsoft Purview DSPM for AI and Endpoint DLP | | [Netskope](/vendors/netskope/) | 3rd | 65.5 | Netskope Skylight AI Command Center, with Netskope Skylight GenAI App Security, Agentic Broker and the Netskope Client endpoint AI discovery, on the Netskope One platform | | [Nudge Security](/vendors/nudge-security/) | 4th | 64.5 | Nudge Security platform: AI and SaaS discovery, AI agent discovery, AI conversation monitoring, browser extension, AI governance nudges | | [Palo Alto Networks](/vendors/palo-alto-networks/) | 7th | 55.5 | AI Access Security, with Prisma Browser | | [Push Security](/vendors/push-security/) | joint 9th | 53.5 | Push browser security platform: AI visibility and control, shadow SaaS and app discovery, browser extension and OAuth inventory | | [Reco](/vendors/reco/) | joint 11th | 49.5 | Reco Shadow AI Discovery / Shadow Agent Discovery, Application Discovery and AI Governance and Security | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 13th | 48.0 | Prompt Security AI Usage Control (SentinelOne) | | [WitnessAI](/vendors/witnessai/) | 16th | 42.0 | WitnessAI Observe (AI application and agent discovery) with Control for policy | | [Zscaler](/vendors/zscaler/) | 6th | 56.5 | Zscaler AI Security: AI Asset Management (Public AI inventory for GenAI, embedded AI, endpoint AI, MCP servers), Endpoint AI Security, and Zscaler Internet Access GenAI controls (Zero Trust Exchange) | ## Related pages - [Ranking of all vendors](/ranking/) - [How vendors are scored](/method/) - [Best vendor by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Aim Security for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/aim-security/ Updated: 2026-11-16 Aim Security's shadow AI product is sold as Cato AI Security for End Users, built on Cato AI Endpoint. Per the vendor, a lightweight client pushed through existing device management, plus inline inspection of AI traffic, finds AI apps, models, local coding agents, MCP servers, skills and Office AI add-ins per user and endpoint. It can block, redact or anonymize prompts, redirect and show pop-up notices. An on-premises or own-cloud option exists. Cato says AI protection goes live in weeks. None found on 16 Nov 2026 for discovery of OAuth-granted AI apps, identity provider or SaaS API sources, a supported operating system list or a published price. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Cato AI Endpoint finds AI apps, local coding agents, MCP servers and skills per user and endpoint, and can block, anonymize or redact prompts (aim-security-10, -11, -23, -24). **Weakness:** None found on 2026-11-16 for OAuth-granted AI apps, identity provider or SaaS API vantage points, supported operating systems, or a published price (aim-security-15, -18, -21, -30). ## Scores and evidence Aim Security is 14th in the overall ranking at 47.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Aim Security: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | AI apps, models, local agents, MCP servers, skills, named coding agents and Office AI add-ins. None found for OAuth apps or named embedded SaaS AI. | [aim-security-10](#fact-aim-security-10), [aim-security-11](#fact-aim-security-11), [aim-security-12](#fact-aim-security-12), [aim-security-14](#fact-aim-security-14), [aim-security-15](#fact-aim-security-15) | | Where it looks | 4 of 10 | 20 | Endpoint client and inline network inspection. None found for IdP, SaaS logs, email or billing. | [aim-security-16](#fact-aim-security-16), [aim-security-12](#fact-aim-security-12), [aim-security-18](#fact-aim-security-18) | | Time to a first inventory | 3 of 10 | 15 | Pushed via device management with protection live in weeks. None found for OS list or time to first inventory. | [aim-security-19](#fact-aim-security-19), [aim-security-20](#fact-aim-security-20), [aim-security-21](#fact-aim-security-21) | | What it can do about a finding | 6 of 10 | 15 | Block, redact or anonymize prompts, redirect and pop-up notices. None found for OAuth revocation, approval or tickets. | [aim-security-22](#fact-aim-security-22), [aim-security-23](#fact-aim-security-23), [aim-security-24](#fact-aim-security-24), [aim-security-25](#fact-aim-security-25) | | Risk context and reporting | 5 of 10 | 10 | Risk scoring, agents by user and endpoint and compliance mapping. None found for a training-on-inputs rating or an export. | [aim-security-26](#fact-aim-security-26), [aim-security-27](#fact-aim-security-27), [aim-security-28](#fact-aim-security-28), [aim-security-29](#fact-aim-security-29) | | Transparency | 2 of 10 | 10 | Published threat research. None found for a price or readable documentation. | [aim-security-32](#fact-aim-security-32), [aim-security-33](#fact-aim-security-33), [aim-security-30](#fact-aim-security-30), [aim-security-31](#fact-aim-security-31) | | Proof | 8 of 10 | 5 | Third parties report $409 million raised, 4,800+ customers and the acquisition; named customer quotes and an IDC quote. | [aim-security-07](#fact-aim-security-07), [aim-security-09](#fact-aim-security-09), [aim-security-08](#fact-aim-security-08), [aim-security-34](#fact-aim-security-34), [aim-security-35](#fact-aim-security-35), [aim-security-36](#fact-aim-security-36) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 14th | 47.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 10th | 51.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 15th | 45.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 13th | 50.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 13th | 49.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 12th | 52.0 | ## Product, deployment, price and owner - Product assessed: Cato AI Security for End Users with Cato AI Endpoint (shadow AI and agent discovery) - Deployment: Lightweight endpoint client pushed through existing device management, plus inline network inspection of AI traffic; proxy, AI gateway or API integration for applications; on-prem or own-cloud Outposts option - Time to deploy: Cato says AI protection goes live in weeks; no operating system list found on 2026-11-16 - Price: none found on 2026-11-16; the product page offers a demo and the /pricing/ page returned 404 - Owner: Cato Networks - Website: [aim.security](https://www.aim.security) ## Facts and sources - **aim-security-01** (vendor): Cato Networks says in its own blog dated 2025-09-03 that it acquired Aim Security, and that Aim would stay available as a standalone product before converging into the Cato platform by early 2026.Quote: "In the near term, Aim will remain available as a standalone product. By early 2026, it will converge into the Cato platform"Source: [catonetworks.com](https://www.catonetworks.com/blog/securing-ai-transformation-why-cato-acquired-aim-security/), read 2026-11-16. - **aim-security-02** (third-party): SecurityWeek reported on 2025-09-03 that Cato Networks acquired Aim Security, which was founded in 2022 and came out of stealth in January 2024.Quote: "Founded in 2022 to help organizations with the secure deployment of generative-AI utilities, Aim emerged from stealth in January 2024."Source: [securityweek.com](https://www.securityweek.com/cato-networks-acquires-ai-security-firm-aim-security/), read 2026-11-16. - **aim-security-03** (third-party): In a PRNewswire release dated 2026-07-27, Cato says its Cato AI Security product combines capabilities from its Aim Security acquisition and was introduced to market in March 2026.Quote: "Cato AI Security, which converges advanced AI governance and protection capabilities from Cato's Aim Security acquisition into a fully integrated solution introduced to market in March 2026"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/cato-networks-exceeds-415m-in-arr-fueled-by-42-yoy-growth-and-enterprise-demand-302833667.html), read 2026-11-16. - **aim-security-04** (third-party): Wikipedia states that Cato Networks announced its first M&A deal, the acquisition of Aim Security, in September 2025.Quote: "Cato Networks announced its first-ever M&A deal with the acquisition of Aim Security, an AI security startup based in Israel"Source: [en.wikipedia.org](https://en.wikipedia.org/wiki/Cato_Networks), read 2026-11-16. - **aim-security-05** (vendor): Cato posts a Bloomberg summary dated 2025-09-07 saying Cato agreed to acquire Aim Security, Cato's first acquisition.Quote: "Israeli cybersecurity company Cato Networks has agreed to acquire AI startup Aim Security, marking the firm’s first acquisition."Source: [catonetworks.com](https://www.catonetworks.com/news/bloomberg-cato-networks-to-buy-aim-security-in-bid-to-make-ai-more-secure/), read 2026-11-16. - **aim-security-06** (none-found): None found on 2026-11-16: https://www.aim.security would not load (HTTP 403 Cloudflare block, also blocked via the r.jina.ai reader and /sitemap.xml), so no Aim-owned page could be read; ownership and product facts come from Cato Networks pages and third-party reports.Source: [aim.security](https://www.aim.security), read 2026-11-16. - **aim-security-07** (third-party): Wikipedia reports Cato Networks raised $50 million more in a Series G, bringing the total to $409 million at an estimated valuation above $4.8 billion.Quote: "it had raised another $50 million in its Series G round, bringing the total to $409 million at an estimated valuation of over $4.8 billion"Source: [en.wikipedia.org](https://en.wikipedia.org/wiki/Cato_Networks), read 2026-11-16. - **aim-security-08** (third-party): SecurityWeek reports Aim raised $10 million in seed funding and $18 million more in June 2025 before the acquisition, and that its researchers found the EchoLeak Microsoft 365 Copilot vulnerability (CVE-2025-32711).Quote: "Aim’s research team recently uncovered the first reported CVE of a zero-click AI vulnerability in Microsoft 365 Copilot"Source: [securityweek.com](https://www.securityweek.com/cato-networks-acquires-ai-security-firm-aim-security/), read 2026-11-16. - **aim-security-09** (third-party): Cato says on 2026-07-27 that it passed $415 million in ARR with 42% year-over-year growth and serves more than 4,800 customers.Quote: "surpassed $415 million in ARR, with 42% year-over-year (YoY) growth"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/cato-networks-exceeds-415m-in-arr-fueled-by-42-yoy-growth-and-enterprise-demand-302833667.html), read 2026-11-16. - **aim-security-10** (vendor): Cato says its AI Security for End Users product inventories sanctioned and unsanctioned AI apps, models and agents across users and sites.Quote: "Cato automatically inventories every sanctioned and unsanctioned AI app, model, and agent in use across every user and site."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-11** (vendor): Cato says AI Endpoint finds local agents and their MCP servers, tools and skills on the endpoint before any traffic reaches Cato.Quote: "Cato finds local agents and their MCP servers, tools and skills on the endpoint, before any traffic reaches Cato, and flags risky ones before they act."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-12** (vendor): Cato says AI Endpoint can find AI applications, local agents, MCP servers and skills on endpoints, and names Claude Code, Codex and Cursor as coding agents it covers.Quote: "Cato AI Endpoint can find the AI applications, local agents, MCP servers and skills on your endpoints"Source: [catonetworks.com](https://www.catonetworks.com/blog/cato-ai-endpoint-secures-browser-ai-desktop-ai-and-coding-agents-on-every-endpoint/), read 2026-11-16. - **aim-security-13** (vendor): Cato lists visibility use cases: discover shadow AI and coding agents on endpoints, inventory LLM and RAG apps, and map agents, MCP servers and skills.Quote: "Discover shadow AI and coding agents across endpoints"Source: [catonetworks.com](https://www.catonetworks.com/solutions/ai-security/), read 2026-11-16. - **aim-security-14** (vendor): Cato says AI Endpoint covers AI add-ins in Office apps, a partial view of AI features embedded in approved software.Quote: "Controls sensitive data in prompts from AI add-ins in Office apps."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-15** (none-found): None found on 2026-11-16: no mention of discovering AI apps granted OAuth access to Google Workspace or Microsoft 365, or of named AI features inside SaaS such as Notion AI or Slack AI, on the Cato AI Security for End Users page.Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-16** (vendor): Cato says AI Endpoint uses a lightweight client for browser AI and desktop AI apps, and also inspects browser and app AI traffic inline on the network with the same engine.Quote: "Browser AI and AI apps are also inspected inline on the network with the same AI engine, so endpoint and network AI events share one dashboard."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-17** (vendor): Cato says its end-user discovery maps usage to users, apps and data without agents.Quote: "Map usage to users, apps, and data"Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-18** (none-found): None found on 2026-11-16: no identity provider integration (Okta, Entra ID, Google), SaaS audit log, email or billing record source is named for AI discovery on the Cato AI Security for End Users page.Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-19** (vendor): Cato says AI Endpoint is pushed out with existing endpoint management tools, runs beside the EDR, and AI protection goes live in weeks.Quote: "Existing SSE, VPN and EDR keep doing their jobs, so AI protection goes live in weeks."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-20** (vendor): Cato says the AI Endpoint client is deployed through device management like any other app, and finding agents needs no network change.Quote: "The Cato AI Endpoint client goes out through device management like any other app."Source: [catonetworks.com](https://www.catonetworks.com/blog/cato-ai-endpoint-secures-browser-ai-desktop-ai-and-coding-agents-on-every-endpoint/), read 2026-11-16. - **aim-security-21** (none-found): None found on 2026-11-16: the Cato AI Endpoint page and launch blog do not list supported operating systems or browsers, or a time to first inventory in hours or days.Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-22** (vendor): Cato lists enforcement actions: block or redact sensitive data, redirect users to approved tools, and log violations for evidence.Quote: "Automatically block or redact sensitive data"Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-23** (vendor): Cato says that when a prompt is sent from a desktop AI app, AI Endpoint checks it and stops a pasted customer list from reaching the AI service, with a pop-up telling the user.Quote: "Cato AI Endpoint checks the prompt and stops the customer list from reaching the AI service."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-24** (vendor): Cato says browser and desktop AI rules can monitor, block or anonymize a prompt, while coding agents get monitor and block controls.Quote: "Coding agents get monitor and block controls, and your team manages browser, desktop and coding agent protection from one dashboard."Source: [catonetworks.com](https://www.catonetworks.com/blog/cato-ai-endpoint-secures-browser-ai-desktop-ai-and-coding-agents-on-every-endpoint/), read 2026-11-16. - **aim-security-25** (vendor): Cato lists "Redirect users to approved tools" as an enforcement option.Quote: "Redirect users to approved tools"Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-26** (vendor): Cato lists risk scoring across sanctioned and shadow AI and usage patterns mapped to security impact.Quote: "Risk scoring across sanctioned and shadow AI"Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-27** (vendor): Cato says AI Endpoint shows each agent by user and endpoint with its MCP servers and skills, listing agents with risky settings first.Quote: "Agents with risky settings, such as unrestricted access to the command line, show up first, so the team knows where to start."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-28** (vendor): Cato says AI Endpoint gives each skill a risk finding with a block, monitor or allow recommendation, and verified remote MCP servers get a badge.Quote: "The Cato AI Endpoint client gives each skill a risk finding with a block, monitor or allow recommendation. Verified remote MCP servers get a badge."Source: [catonetworks.com](https://www.catonetworks.com/blog/cato-ai-endpoint-secures-browser-ai-desktop-ai-and-coding-agents-on-every-endpoint/), read 2026-11-16. - **aim-security-29** (vendor): Cato says it maps AI usage to standards such as EU AI Act, GDPR, NIST and OWASP for auditable evidence.Quote: "Map AI usage to corporate guidelines and standards like EU AI Act, GDPR, NIST, and OWASP."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-30** (none-found): None found on 2026-11-16: no published price for Cato AI Security or Cato AI Endpoint on the product page, which offers "Get a product demo"; https://www.catonetworks.com/pricing/ returned HTTP 404.Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-31** (none-found): None found on 2026-11-16: the Cato support and documentation site at support.catonetworks.com returned HTTP 403 to an automated request, so public product documentation could not be read.Source: [support.catonetworks.com](https://support.catonetworks.com/), read 2026-11-16. - **aim-security-32** (vendor): Cato publishes threat research on its public blog; a 2025-04-15 Cato CTRL post covers risks of the Model Context Protocol and security practices.Quote: "As MCP-enabled tools interact with sensitive data, they increase the risk of unauthorized access."Source: [catonetworks.com](https://www.catonetworks.com/blog/cato-ctrl-exploiting-model-context-protocol-mcp/), read 2026-11-16. - **aim-security-33** (vendor): Cato offers public product pages, a white paper and a webinar on agentic AI security.Quote: "Securing agentic AI white paper"Source: [catonetworks.com](https://www.catonetworks.com/platform/agentic-ai-security/), read 2026-11-16. - **aim-security-34** (vendor): Cato publishes a quote from Nithin Reddy, VP Security at Dayforce, who says Dayforce deployed Cato AI extensions and put Cato AI security inline.Quote: "We prioritized the AI our employees use, so we deployed Cato AI extensions for visibility and to prevent prompt injection and data leakage."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-35** (vendor): Cato publishes a quote from Doug Innocenti, CIO/CSO at MoonPay, on using Cato AI Security.Quote: "Having the right AI security solution gave me the confidence to say yes to the business."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-applications/), read 2026-11-16. - **aim-security-36** (vendor): Cato's 2026-08-03 launch release for Agentic Threat Prevention quotes Frank Dickson, group vice president at IDC, on AI-assisted attacks.Quote: "An entire attack can require less time than is required to investigate an alert"Source: [catonetworks.com](https://www.catonetworks.com/news/cato-launches-agentic-threat-prevention-to-predict-mitigate-ai-assisted-attacks/), read 2026-11-16. - **aim-security-37** (vendor): On 2026-10-06 Cato said it earned the AWS Security Competency in AI Security (Agentic AI category), one of fewer than 25 AWS partners to do so.Quote: "Cato is one of fewer than 25 AWS partners to earn the AI Security Competency"Source: [catonetworks.com](https://www.catonetworks.com/news/cato-achieves-aws-security-competency-in-ai-security/), read 2026-11-16. - **aim-security-38** (vendor): On 2026-10-27 Cato launched Cato AI Endpoint; the release quotes Alfred Sobrinho, CIO of PlayPower, and Chris DePuy of 650 Group.Quote: "“AI is creating real productivity gains across our business, but it also introduces new concerns about visibility, data protection, and governance,”"Source: [catonetworks.com](https://www.catonetworks.com/news/cato-launches-cato-ai-endpoint-secure-ai-use-beyond-browser/), read 2026-11-16. - **aim-security-39** (vendor): On 2026-10-27 Cato said Cato AI Endpoint is a client that discovers AI use, protects data and enforces policy across browsers, desktop applications and local agents.Quote: "a new endpoint client that helps organizations discover AI use, protect sensitive data, and enforce AI policies across browsers, desktop applications, and local agents"Source: [catonetworks.com](https://www.catonetworks.com/news/cato-launches-cato-ai-endpoint-secure-ai-use-beyond-browser/), read 2026-11-16. - **aim-security-42** (vendor): Cato says in the browser the same rules can monitor, anonymize or block a prompt.Quote: "In the browser, the same rules can monitor a prompt, anonymize the sensitive data in the prompt, or block the prompt."Source: [catonetworks.com](https://www.catonetworks.com/platform/cato-ai-endpoint/), read 2026-11-16. - **aim-security-43** (vendor): Cato says its extensions give visibility into employee AI use and guard against prompt injection and leakage (customer Dayforce).Quote: "we deployed Cato AI extensions for visibility and to prevent prompt injection and data leakage."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-end-users/), read 2026-11-16. - **aim-security-49** (vendor): Cato says AI Security for Applications deploys via proxy interception, AI-gateway integrations or APIs into model platforms and developer tools.Quote: "Deploy via proxy interception, AI-gateway integrations, or APIs into model platforms and developer tools."Source: [catonetworks.com](https://www.catonetworks.com/platform/ai-security-for-applications/), read 2026-11-16. - **aim-security-50** (vendor): Cato says applications are protected through the Cato cloud, the customer's existing AI gateway, or on-prem with Outposts.Quote: "Protects apps through the Cato cloud, the AI gateway you already run, or on-prem with Outposts."Source: [catonetworks.com](https://www.catonetworks.com/solutions/ai-security/), read 2026-11-16. - **aim-security-70** (vendor): Cato says AI Security for Applications can run on-prem or in the customer cloud with AI Security Outposts so prompts stay in the customer environment.Quote: "Run AI Security for Applications on-prem or in your own cloud with AI Security Outposts. Prompts stay in your environment."Source: [catonetworks.com](https://www.catonetworks.com/solutions/ai-security/), read 2026-11-16. - **aim-security-72** (vendor): Dated news: on 2026-08-03 Cato launched Agentic Threat Prevention, an autonomous-agent capability to predict attack paths; it is network threat prevention, not AI runtime control.Quote: "today announced the launch of Cato Agentic Threat Prevention, a new capability that deploys autonomous agents to predict likely attack paths"Source: [catonetworks.com](https://www.catonetworks.com/news/cato-launches-agentic-threat-prevention-to-predict-mitigate-ai-assisted-attacks/), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Zscaler profile](/vendors/zscaler/) - [Aurascape profile](/vendors/aurascape/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Aurascape for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/aurascape/ Updated: 2026-11-16 Aurascape discovers AI through an inline, AI-aware proxy that steers only AI traffic and runs alongside an existing SASE stack. Its pages describe a catalog of more than 20,000 AI apps, embedded SaaS AI, agents, IDE plugins and shadow MCP servers, with more than 25 risk attributes per app, including training and retention. Actions include block, action-level block, warn, redact, coach and time-limited exception requests. It lists a 48-hour service level for connectors to new AI apps. None found on 16 Nov 2026 for a price, a time to a first inventory, an operating system or browser list, OAuth grant discovery or revocation, or a named identity provider integration. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Broad documented discovery of AI apps (20,000+ catalogued), embedded SaaS AI, agents, IDE plugins and shadow MCP servers, with per-app risk scoring and block, coach, redact and mask actions (aurascape-07, aurascape-08, aurascape-09, aurascape-21, aurascape-24). **Weakness:** No price, no time to first inventory, no OS or browser list, no OAuth grant discovery or revocation and no named identity provider integration found on 2026-11-16 (aurascape-11, aurascape-17, aurascape-20, aurascape-35). ## Scores and evidence Aurascape is joint 9th in the overall ranking at 53.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Aurascape: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 7 of 10 | 25 | Long-tail AI apps, agents, embedded SaaS AI, MCP servers, IDE plugins, extensions and named coding assistants, vendor-only. None found for OAuth-granted AI apps. | [aurascape-07](#fact-aurascape-07), [aurascape-08](#fact-aurascape-08), [aurascape-09](#fact-aurascape-09), [aurascape-10](#fact-aurascape-10), [aurascape-11](#fact-aurascape-11) | | Where it looks | 3 of 10 | 20 | Inline AI proxy traffic including thick clients, plus file repository connections. None found for IdP, endpoint, browser extension, email or billing. | [aurascape-12](#fact-aurascape-12), [aurascape-13](#fact-aurascape-13), [aurascape-15](#fact-aurascape-15), [aurascape-16](#fact-aurascape-16), [aurascape-17](#fact-aurascape-17) | | Time to a first inventory | 3 of 10 | 15 | Inline steering with no PAC files per a case study. None found for a time to first inventory or OS and browser coverage. | [aurascape-14](#fact-aurascape-14), [aurascape-19](#fact-aurascape-19), [aurascape-18](#fact-aurascape-18), [aurascape-20](#fact-aurascape-20) | | What it can do about a finding | 7 of 10 | 15 | Block, action-level block, warn or confirm, redact, coach and time-limited exception requests. None found for OAuth revocation or tickets. | [aurascape-21](#fact-aurascape-21), [aurascape-22](#fact-aurascape-22), [aurascape-23](#fact-aurascape-23), [aurascape-24](#fact-aurascape-24), [aurascape-25](#fact-aurascape-25), [aurascape-26](#fact-aurascape-26) | | Risk context and reporting | 7 of 10 | 10 | 25+ risk attributes per app including training and retention, account type, and reports by user and app; vendor-only. | [aurascape-27](#fact-aurascape-27), [aurascape-28](#fact-aurascape-28), [aurascape-29](#fact-aurascape-29), [aurascape-30](#fact-aurascape-30), [aurascape-31](#fact-aurascape-31) | | Transparency | 5 of 10 | 10 | Public FAQ, a sample report and research posts. None found for a price. | [aurascape-32](#fact-aurascape-32), [aurascape-33](#fact-aurascape-33), [aurascape-34](#fact-aurascape-34), [aurascape-35](#fact-aurascape-35) | | Proof | 6 of 10 | 5 | $50 million funding reported by SiliconANGLE, named customers and a Gartner Hype Cycle mention claimed by the vendor. | [aurascape-03](#fact-aurascape-03), [aurascape-04](#fact-aurascape-04), [aurascape-36](#fact-aurascape-36), [aurascape-37](#fact-aurascape-37), [aurascape-39](#fact-aurascape-39) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | joint 9th | 53.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 7th | 58.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 12th | 48.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 9th | 54.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 10th | 52.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 6th | 58.5 | ## Product, deployment, price and owner - Product assessed: Aurascape AI Security Platform, Discover and Monitor AI and Control AI Access and Usage (AI Proxy based discovery and risk scoring) - Deployment: Inline AI-aware proxy that steers only AI traffic, run alongside an existing SASE stack, with no PAC files or local routing changes (per customer case studies) - Time to deploy: none found on 2026-11-16 (only a 48-hour connector SLA for new AI apps and a gateway said to deploy in minutes) - Price: none found on 2026-11-16 - Owner: Aurascape, Inc. (standalone, venture-backed; no parent company named on its pages) - Website: [aurascape.ai](https://aurascape.ai) ## Facts and sources - **aurascape-01** (vendor): Aurascape's own company page lists Moinul Khan as CEO and co-founder and describes the company as founded by veterans of Palo Alto Networks, Zscaler, FireEye, Juniper and Google.Quote: "Our team led the way at Palo Alto Networks, Zscaler, FireEye, Juniper, and Google"Source: [aurascape.ai](https://aurascape.ai/company/), read 2026-11-16. - **aurascape-02** (vendor): The Aurascape terms of service say the site is provided by Aurascape, Inc., a standalone company with no parent named.Quote: "The Site is provided by Aurascape, Inc. (“Aurascape”, “our” or “us”)"Source: [aurascape.ai](https://aurascape.ai/terms-of-service/), read 2026-11-16. - **aurascape-03** (third-party): SiliconANGLE reported on 2025-04-08 that Aurascape was founded in 2023 and launched with $50 million in funding led by Menlo Ventures and Mayfield Fund.Quote: "Lead investors in the company include Menlo Ventures and Mayfield Fund"Source: [siliconangle.com](https://siliconangle.com/2025/04/08/aurascape-launches-50m-funding-bring-security-observability-ai-apps/), read 2026-11-16. - **aurascape-04** (third-party): Ventureburn, in a piece dated 8 May 2026, repeats that Aurascape raised $50 million from investors including Menlo Ventures and Mayfield Fund.Quote: "Aurascape raised $50 million in funding from investors, including Menlo Ventures and Mayfield Fund."Source: [ventureburn.com](https://ventureburn.com/aurascape-raises-50m-to-secure-enterprise-ai-applications/), read 2026-11-16. - **aurascape-05** (vendor): Aurascape's product page names the platform as one product covering AI access and usage control, data protection, and securing AI apps and agents.Quote: "One platform to secure how your business uses and builds AI."Source: [aurascape.ai](https://aurascape.ai/product/), read 2026-11-16. - **aurascape-06** (vendor): Aurascape says it specializes in finding the long tail of new AI apps and continuously adds them to the inventory.The exact quote is on the source page.Source: [aurascape.ai](https://aurascape.ai/discover-and-monitor-ai/), read 2026-11-16. - **aurascape-07** (vendor): Aurascape says it catalogs 20,000+ AI applications as of 2026-06-24, with identification, risk assessment and entitlement tier decoding for each.Quote: "currently containing 20,000+ catalogued applications as of June 24th, 2026"Source: [aurascape.ai](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/), read 2026-11-16. - **aurascape-08** (vendor): Aurascape says it finds employee-used agents, agents embedded in SaaS, internally built agents and shadow MCP servers.Quote: "Find employee-used agents, agents embedded in SaaS, internally built agents, and shadow MCP servers connected without IT oversight."Source: [aurascape.ai](https://aurascape.ai/secure-agentic-ai/), read 2026-11-16. - **aurascape-09** (vendor): Aurascape says it discovers risky IDE plugins, assistant extensions and MCP servers in development environments and supports coding assistants such as Claude Code, Cursor, GitHub Copilot, Codex and Windsurf.Quote: "Risky IDE plugins, assistant extensions, and MCP servers are discovered, so security can govern which ones are used with enterprise code."Source: [aurascape.ai](https://aurascape.ai/coding-assistant-guardrails/), read 2026-11-16. - **aurascape-10** (vendor): Aurascape says it finds AI embedded inside trusted SaaS apps and websites, a case where traditional tools only see the destination.Quote: "Patented discovery finds known and long-tail AI apps and agents, including embedded AI hiding inside trusted SaaS and websites."Source: [aurascape.ai](https://aurascape.ai/see-and-control-ai-security-platform/), read 2026-11-16. - **aurascape-11** (none-found): None found on 2026-11-16: looked for discovery of AI apps granted OAuth access to Google Workspace or Microsoft 365 on the discovery, FAQ, platform and agentic AI pages; the only OAuth mention is OAuth 2.1 authorization for its own MCP gateway.Source: [aurascape.ai](https://aurascape.ai/discover-and-monitor-ai/), read 2026-11-16. - **aurascape-12** (vendor): Aurascape says its discovery works by inspecting and categorizing AI traffic in motion, reading queries and responses.Quote: "Aurascape inspects and categorizes traffic in motion, with hundreds of data categories, themes, and topics"Source: [aurascape.ai](https://aurascape.ai/discover-and-monitor-ai/), read 2026-11-16. - **aurascape-13** (vendor): A published customer case study says Aurascape runs alongside an existing SASE stack and steers only AI traffic, inline.Quote: "Aurascape runs alongside the firm’s existing SASE stack, steering only AI traffic and decoding and governing it inline."Source: [aurascape.ai](https://aurascape.ai/resources/case-study/securing-ai-coding-assistants-financial-services/), read 2026-11-16. - **aurascape-14** (vendor): A customer case study says Aurascape protection is inline with no PAC files or local routing changes, steering only AI traffic.Quote: "Inline, real-time prevention with no PAC files or local routing changes."Source: [aurascape.ai](https://aurascape.ai/resources/case-study/the-police-credit-union-ai-compliance-case-study/), read 2026-11-16. - **aurascape-15** (vendor): Aurascape says it decodes thick-client, command-line and modern protocol traffic alongside the browser, so coding assistants outside the browser are visible.Quote: "Thick-client, CLI, and modern protocol traffic decoded alongside the browser, so the coding interaction is visible wherever it happens."Source: [aurascape.ai](https://aurascape.ai/see-and-control-ai-security-platform/), read 2026-11-16. - **aurascape-16** (vendor): Aurascape says its Copilot Readiness feature connects to file repositories to find misconfigured permissions and oversharing.Quote: "Aurascape connects and traverses through file repositories to make sure you’re ready to grant Copilots only appropriate access."Source: [aurascape.ai](https://aurascape.ai/copilot-readiness/), read 2026-11-16. - **aurascape-17** (none-found): None found on 2026-11-16: looked for a named identity provider (Okta, Entra ID, Google), endpoint agent, browser extension, or billing or email record integration for AI discovery on the product, discovery, FAQ and case study pages; the sample report mentions active devices but no named integration.Source: [aurascape.ai](https://aurascape.ai/product/), read 2026-11-16. - **aurascape-18** (vendor): Aurascape says it builds production connectors for most new AI apps within 48 hours.Quote: "48-hour production deployment for the vast majority of AI applications, with strong coverage across the AI application landscape."Source: [aurascape.ai](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/), read 2026-11-16. - **aurascape-19** (vendor): A customer case study describes a two-phase rollout, visibility first to build the AI inventory, then protection.Quote: "Deploy Aurascape in two phases: visibility to build the AI inventory and risk picture, then protection"Source: [aurascape.ai](https://aurascape.ai/resources/case-study/the-police-credit-union-ai-compliance-case-study/), read 2026-11-16. - **aurascape-20** (none-found): None found on 2026-11-16: looked for a published time to first AI inventory, supported operating systems and browsers, and whether an agent or extension is required, on the product, discovery, FAQ and case study pages.Source: [aurascape.ai](https://aurascape.ai/discover-and-monitor-ai/), read 2026-11-16. - **aurascape-21** (vendor): Aurascape says policy can block a risky AI tool entirely or only unsafe actions inside less risky tools.Quote: "Aurascape can entirely block access to risky AI tools, and only unsafe actions within less risky AI applications."Source: [aurascape.ai](https://aurascape.ai/frictionless-ai-security/), read 2026-11-16. - **aurascape-22** (vendor): Aurascape says it explains blocks to the user, offers safer actions, and lets users request limited-time exceptions.Quote: "can give users the option to request limited-time-exceptions if they disagree with the block"Source: [aurascape.ai](https://aurascape.ai/frictionless-ai-security/), read 2026-11-16. - **aurascape-23** (vendor): Aurascape says it can redact sensitive data and coach users toward the enterprise tenant instead of blocking.Quote: "Redact the account numbers and let the request through. Coach the user toward the enterprise tenant."Source: [aurascape.ai](https://aurascape.ai/see-and-control-ai-security-platform/), read 2026-11-16. - **aurascape-24** (vendor): On 2026-10-07 Aurascape described its September 2026 release, which adds masking of sensitive values in prompts before they reach the model.Quote: "masking replaces a sensitive value in a prompt before it reaches the model."Source: [aurascape.ai](https://aurascape.ai/resources/blog/september-2026-product-update/), read 2026-11-16. - **aurascape-25** (vendor): In the same 2026-10-07 post Aurascape says model-level policies can warn, ask for confirmation, or block use of a restricted model across services such as Claude, ChatGPT and Amazon Bedrock.Quote: "Access policies can coach users with warning or confirmation notifications, or block activity when they use a model restricted by the organization."Source: [aurascape.ai](https://aurascape.ai/resources/blog/september-2026-product-update/), read 2026-11-16. - **aurascape-26** (none-found): None found on 2026-11-16: looked for revoking OAuth grants or routing a discovered AI tool to a ticketing system on the discovery, FAQ, frictionless and governance pages.Source: [aurascape.ai](https://aurascape.ai/frictionless-ai-security/), read 2026-11-16. - **aurascape-27** (vendor): Aurascape says its risk analysis covers terms and conditions, authentication and entitlement, and how data is handled.Quote: "This risk analysis considers all known information and vulnerabilities for the tool, the most up-to-date T’s and C’s, user authentication and entitlement, and exactly how your data is handled."Source: [aurascape.ai](https://aurascape.ai/frictionless-ai-security/), read 2026-11-16. - **aurascape-28** (vendor): Aurascape says its risk bot collects 25+ attributes per app including data training, retention and generated content ownership.Quote: "Automated collection of 25+ risk attributes"Source: [aurascape.ai](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/), read 2026-11-16. - **aurascape-29** (vendor): Aurascape says it shows whether users reach AI tools with free personal accounts or paid enterprise seats.Quote: "see if users access AI tools with free, personal accounts versus paid, enterprise seats"Source: [aurascape.ai](https://aurascape.ai/discover-and-monitor-ai/), read 2026-11-16. - **aurascape-30** (vendor): Aurascape publishes a sample AI security report that counts conversations, active users and active devices, ranks applications and users, and flags newly observed unsanctioned tools.Quote: "It counts AI conversations, active users, and active devices, then ranks the applications and people driving the most activity."Source: [aurascape.ai](https://aurascape.ai/resources/sample-report/sample-ai-security-report/), read 2026-11-16. - **aurascape-31** (vendor): On 2026-10-07 Aurascape said on-demand reports add Unique User Counts by Application and a Top Protection Policy Hits chart.Quote: "now include Unique User Counts by Application, showing how many distinct users were active on each application during the reporting period."Source: [aurascape.ai](https://aurascape.ai/resources/blog/september-2026-product-update/), read 2026-11-16. - **aurascape-32** (vendor): Aurascape publishes a long FAQ and solution pages readable without sign-in.Quote: "An AI security platform is a system designed to monitor, control, and protect AI usage and AI development."Source: [aurascape.ai](https://aurascape.ai/ai-security-platform-faq/), read 2026-11-16. - **aurascape-33** (vendor): Aurascape publishes a sample AI security report and states its figures are illustrative.Quote: "The sample uses illustrative figures modeled on patterns from real Aurascape deployments."Source: [aurascape.ai](https://aurascape.ai/resources/sample-report/sample-ai-security-report/), read 2026-11-16. - **aurascape-34** (vendor): Aurascape lists research and technical posts such as an Aura Labs report on a ChatGPT agent mode vulnerability, dated 2025-08-26, in its public posts index.Quote: "Aurascape Aura Labs Identifies and Reports ChatGPT Agent Mode Vulnerability"Source: [aurascape.ai](https://aurascape.ai/wp-json/wp/v2/posts?per_page=100&_fields=date,link,title), read 2026-11-16. - **aurascape-35** (none-found): None found on 2026-11-16: looked for a published price or pricing page; the site page list and the product page show only Book a Demo calls to action, and a request to /pricing returned 403.Source: [aurascape.ai](https://aurascape.ai/product/), read 2026-11-16. - **aurascape-36** (vendor): Aurascape's homepage carries a named customer quote from the CTO of WinWire Technologies.Quote: "Aurascape gives us full visibility and protection for risky AI use"Source: [aurascape.ai](https://aurascape.ai/), read 2026-11-16. - **aurascape-37** (vendor): Aurascape publishes a case study naming The Police Credit Union as a customer.Quote: "The Police Credit Union (TPCU) deployed Aurascape in two phases, visibility first and protection second"Source: [aurascape.ai](https://aurascape.ai/resources/case-study/the-police-credit-union-ai-compliance-case-study/), read 2026-11-16. - **aurascape-38** (vendor): An unnamed Fortune 500 healthcare technology customer reports unsanctioned AI access fell to near zero, per Aurascape's case study.Quote: "Unsanctioned AI access fell to near zero while governed AI use grew to 60,000+ users worldwide."Source: [aurascape.ai](https://aurascape.ai/resources/case-study/unsanctioned-ai-near-zero-case-study/), read 2026-11-16. - **aurascape-39** (vendor): On 2026-08-20 Aurascape announced it was recognized in the Gartner AI Governance Technologies Hype Cycle (title read from its posts index; the page body was not readable).Quote: "Aurascape Recognized in the Gartner® AI Governance Technologies Hype Cycle®"Source: [aurascape.ai](https://aurascape.ai/wp-json/wp/v2/posts?per_page=100&_fields=date,link,title), read 2026-11-16. - **aurascape-40** (vendor): Aurascape's company page quotes Richard Stiennon of IT-Harvest on its approach.Quote: "The AI-native, prevention-first approach Aurascape takes makes them stand out in a crowded space."Source: [aurascape.ai](https://aurascape.ai/company/), read 2026-11-16. - **aurascape-41** (vendor): On 2026-08-03 Aurascape named Jim Walsh Chief Revenue Officer and said demand comes from Fortune 500 companies in regulated industries.Quote: "driven by demand from Fortune 500 companies in highly regulated industries"Source: [aurascape.ai](https://aurascape.ai/resources/news/aurascape-jim-walsh-chief-revenue-officer/), read 2026-11-16. ## News mentioning Aurascape - [Aurascape adds masking and model-level policy](/news/aurascape-september-release-masking-and-model-level-policy/) ## Related pages - [All vendor profiles](/vendors/) - [Aim Security profile](/vendors/aim-security/) - [Bay profile](/vendors/bay/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Bay for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/bay/ Updated: 2026-11-16 Bay Shadow AI Discovery, also called AI Asset Discovery, runs from one script pushed through an existing MDM or EDR tool, which launches an ephemeral binary. Bay says an inventory appears in under 5 minutes, covering agents, MCP servers, extensions, command-line tools, models and browser chatbots across devices, browsers and cloud workloads, with a 1-to-5 autonomy score and malicious-indicator scanning. Actions named are allow, warn, steer or block per team, plus domain and extension blocking and an approval workflow. None found on 16 Nov 2026 for OAuth-granted AI apps, or for network, SaaS API or billing sources. Bay is a young company that left stealth in 2026, publishes no price and has little independent testing behind its claims. Bay is a young company that left stealth in 2026, publishes no price and has little independent testing behind its claims. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Discovery spans devices, browsers and cloud workloads with a 1-to-5 autonomy score and malicious-indicator scanning (bay-11, bay-28, bay-26). **Weakness:** None found on 2026-11-16 for OAuth-granted AI apps, network, SaaS API or billing vantage points, or an exportable report (bay-32, bay-33). ## Scores and evidence Bay is joint 11th in the overall ranking at 49.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Bay: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | Vendor claims cover agents, MCP servers, extensions, CLI tools, models, browser chatbots and AI embedded in SaaS. No second source; none found for OAuth-granted AI apps. | [bay-11](#fact-bay-11), [bay-13](#fact-bay-13), [bay-14](#fact-bay-14), [bay-15](#fact-bay-15), [bay-32](#fact-bay-32) | | Where it looks | 4 of 10 | 20 | Endpoint, browser and an identity provider connection are claimed by the vendor. None found for network, SaaS API, email or billing vantage points. | [bay-11](#fact-bay-11), [bay-12](#fact-bay-12), [bay-17](#fact-bay-17), [bay-30](#fact-bay-30), [bay-33](#fact-bay-33) | | Time to a first inventory | 7 of 10 | 15 | Vendor publishes an inventory in under 5 minutes via one MDM or EDR script; SiliconANGLE and SecurityWeek only repeat the vendor's claim. None found for a supported OS or browser list. | [bay-18](#fact-bay-18), [bay-16](#fact-bay-16), [bay-19](#fact-bay-19), [bay-20](#fact-bay-20), [bay-21](#fact-bay-21) | | What it can do about a finding | 6 of 10 | 15 | Allow, warn, steer or block per team, domain and extension blocking, and an approval workflow without published detail. None found for OAuth revocation or ticket routing. | [bay-23](#fact-bay-23), [bay-24](#fact-bay-24), [bay-25](#fact-bay-25), [bay-32](#fact-bay-32) | | Risk context and reporting | 4 of 10 | 10 | Per-tool autonomy rating, malicious-indicator vetting and a user-to-agent graph are claimed. None found for a data-use or training rating or an exportable report. | [bay-28](#fact-bay-28), [bay-26](#fact-bay-26), [bay-29](#fact-bay-29), [bay-33](#fact-bay-33) | | Transparency | 0 of 10 | 10 | Documentation is observed behind sign-in; none found for a price or published research. | [bay-34](#fact-bay-34), [bay-35](#fact-bay-35), [bay-36](#fact-bay-36) | | Proof | 6 of 10 | 5 | Third-party reports confirm $50 million in funding and name customers, with TNW noting unverified claims. None found for analyst coverage. | [bay-06](#fact-bay-06), [bay-07](#fact-bay-07), [bay-08](#fact-bay-08), [bay-09](#fact-bay-09) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | joint 11th | 49.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 13th | 48.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | joint 6th | 55.0 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 10th | 52.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 12th | 50.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 11th | 53.0 | ## Product, deployment, price and owner - Product assessed: Bay Shadow AI Discovery (AI Asset Discovery) - Deployment: Agentless: one script through an existing MDM or EDR runs Wave, an ephemeral binary; Bay Enforcer, a static binary, decides on the device via native hook; no proxy (vendor claims) - Time to deploy: Under 5 minutes, as claimed by the vendor; proof of value in under 14 days (SecurityWeek quote) - Price: none found on 2026-11-16 - Owner: Bay AI, Inc. - Website: [bay.io](https://bay.io/?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor) ## Facts and sources - **bay-01** (vendor): Bay AI, Inc. publishes the Bay site and states its copyright on every page.Quote: "© 2026 Bay AI, Inc. All rights reserved."Source: [bay.io](https://www.bay.io/about?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-02** (vendor): Bay says it was founded by three engineers from endpoint and API security.Quote: "Bay was founded by three engineers who spent their careers solving the hardest problems in endpoint and API security."Source: [bay.io](https://www.bay.io/about?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-03** (third-party): On 2026-11-10 Bay named its leadership: Yuval Zacharia is CEO and co-founder, Aner Morag is CPO and Boris Danilovich is VP of R&D, per SiliconANGLE.Quote: "Zacharia founded Bay with Chief Product Officer Aner Morag and Vice President of Research and Development Boris Danilovich."Source: [siliconangle.com](https://siliconangle.com/2026/11/10/ai-security-startup-bay-launches-with-50m-to-vet-agent-actions-before-they-run/), read 2026-11-16. - **bay-04** (vendor): Bay says it was founded in 2026 and is headquartered in Tel Aviv; its press release is dated 2026-11-10.Quote: "Founded in 2026, Bay is headquartered in Tel Aviv and serves customers in the United States and Europe."Source: [bay.io](https://www.bay.io/blog/bay-emerges-from-stealth-with-50m-after-fortune-100-companies-used-its-platform-to-triple-autonomous-ai-agent-use-2?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-05** (vendor): On 2026-11-10 Bay said it had raised $50 million in total funding.Quote: "Bay has raised $50 million in total funding."Source: [bay.io](https://www.bay.io/blog/bay-emerges-from-stealth-with-50m-after-fortune-100-companies-used-its-platform-to-triple-autonomous-ai-agent-use-2?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-06** (third-party): On 2026-11-11 SecurityWeek reported Bay emerged from stealth with a $37 million Series A led by Battery Ventures plus a $13 million seed.Quote: "emerged from stealth with $50 million funding (comprising a $37 million Series A round led by Battery Ventures"Source: [securityweek.com](https://www.securityweek.com/bay-raises-50-million-to-control-what-enterprise-ai-agents-can-do/), read 2026-11-16. - **bay-07** (third-party): On 2026-11-10 SiliconANGLE reported Micron, Wix, ClickUp and Peloton are among dozens of Bay enterprise customers.Quote: "Micron Technology Inc., Wix.com Ltd., ClickUp Inc. and Peloton Interactive Inc. are among the dozens of enterprise customers signed in the nine months since Bay was founded."Source: [siliconangle.com](https://siliconangle.com/2026/11/10/ai-security-startup-bay-launches-with-50m-to-vet-agent-actions-before-they-run/), read 2026-11-16. - **bay-08** (third-party): The Next Web noted that Bay did not name its Fortune 100 customers or say how it measured the tripling claim, and that one named customer is also an investor.Quote: "The company did not name those customers or say how it measured the figure, and TNW has not verified it."Source: [thenextweb.com](https://thenextweb.com/news/bay-50m-ai-agent-security-stealth-battery-ventures), read 2026-11-16. - **bay-09** (none-found): None found on 2026-11-16: independent analyst coverage of Bay (Gartner, Forrester or similar) on the Bay home page or blog.Source: [bay.io](https://bay.io/?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-10** (vendor): Bay says it is SOC 2 and ISO 27001 certified.Quote: "SOC 2 & ISO 27001 Certified"Source: [bay.io](https://www.bay.io/about?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-11** (vendor): Bay says shadow AI discovery covers agents, models, MCP servers, extensions and skills plus AI built into approved tools, across devices, browsers and cloud workloads.Quote: "Every agent, model, MCP server, extension and skill, plus the AI built into tools your teams already use, across devices, browsers and cloud workloads."Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-12** (vendor): Bay says its browser coverage finds AI tools, extensions and accounts in every browser.Quote: "Find every AI tool, extension and account in every browser, and keep company data out of personal AI."Source: [bay.io](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-13** (vendor): Bay says browser discovery covers chatbots, extensions and embedded SaaS agents, and separates corporate from personal accounts.Quote: "and embedded SaaS agent, from ChatGPT to Gemini"Source: [bay.io](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-14** (vendor): Bay says its endpoint product finds agents, extensions, CLI tools, MCP servers, skills and models on employee devices.Quote: "Bay finds every agent, extension, CLI tool, MCP server, skill and model on employee devices."Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-15** (vendor): Bay says its first scan maps agents, MCP servers, extensions, models and skills on devices, including unapproved ones.Quote: "On the first scan, Bay maps every agent, MCP server, extension, model and skill on devices, including ones nobody approved."Source: [bay.io](https://www.bay.io/platform?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-16** (vendor): Bay says it deploys with one script through an existing MDM, with no proxy and nothing installed by users.Quote: "Bay deploys with one script through an existing MDM. It does not require a proxy to route traffic through and does not require anything to be installed by users."Source: [bay.io](https://www.bay.io/platform?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-17** (vendor): Bay says it is agentless and deploys via MDM or EDR using Wave, an ephemeral binary, with no persistent agent and no proxy.Quote: "Bay is agentless: it deploys through your existing MDM or EDR using Wave, an ephemeral binary, so there’s no persistent agent to maintain and no proxy in the path."Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-18** (vendor): Bay says the endpoint inventory appears in under 5 minutes.Quote: "Bay brings it all into view in under 5 minutes and keeps it in bounds, without slowing anyone down."Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-19** (third-party): On 2026-11-10 SiliconANGLE reported Bay installs through the customer's existing MDM or EDR and the company says it takes under five minutes.Quote: "Bay installs through the mobile device management or endpoint detection and response tools a customer already has, and the company says it takes under five minutes."Source: [siliconangle.com](https://siliconangle.com/2026/11/10/ai-security-startup-bay-launches-with-50m-to-vet-agent-actions-before-they-run/), read 2026-11-16. - **bay-20** (third-party): On 2026-11-11 SecurityWeek quoted Bay saying a fleet deploys in under five minutes and a proof of value completes in under 14 days.Quote: "complete a proof of value in less than 14 days."Source: [securityweek.com](https://www.securityweek.com/bay-raises-50-million-to-control-what-enterprise-ai-agents-can-do/), read 2026-11-16. - **bay-21** (none-found): None found on 2026-11-16: a list of supported operating systems for the Bay endpoint scanner, on the endpoint product page and the platform page.Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-22** (vendor): Bay says it identifies personal versus corporate AI accounts including in personal Chrome profiles; no full browser list was found.Quote: "including in personal Chrome profiles"Source: [bay.io](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-23** (vendor): Bay says browser guardrails keep personal data, secrets and source code out of prompts, with allow, warn, steer or block per team.Quote: "Keep personal data, secrets and source code out of prompts, and decide what to, Allow, Warn, Steer or Block, tailored to each team."Source: [bay.io](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-24** (vendor): Bay says it can allow or block specific AI domains and extensions and stop risky agentic actions inside the browser.Quote: "Bay lets you allow or block specific AI domains and extensions, and stops risky agentic actions inside the browser"Source: [bay.io](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-25** (vendor): Bay lists an approval workflow for newly found AI tools; details of the workflow were not published on the page.Quote: "An approval workflow people actually use"Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-26** (vendor): Bay says every plugin, package, skill and MCP server is scanned and vetted before approval and each time it changes.Quote: "Every plugin, package, skill and MCP server is scanned and vetted before it’s approved and every time it changes."Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-27** (vendor): Bay says every agent, package, extension, MCP server and skill is scanned for malicious indicators before approval.Quote: "Every agent, package, extension, MCP server and skill is scanned for malicious indicators before its approved."Source: [bay.io](https://www.bay.io/platform?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-28** (vendor): Bay says it rates each AI tool from 1 to 5 by how freely it can act on its own.Quote: "Bay rates each tool from 1 to 5 by how freely it can act on its own, so you know where to look first."Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-29** (vendor): Bay says it builds an entity graph of the user behind each agent, the MCP servers and SaaS apps it touches, and the access it holds.Quote: "Bay builds a contextual entity graph of how everything connects: the user behind each agent, the MCP servers and SaaS apps it touches, and the access it holds."Source: [bay.io](https://www.bay.io/platform?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-30** (vendor): Bay says it connects to the identity provider so policies apply to the right people and groups.Quote: "Bay connects to your identity provider, so every policy applies to the right people and groups, and keeps up as your organization changes."Source: [bay.io](https://www.bay.io/ai-governance-compliance?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-32** (none-found): None found on 2026-11-16: discovery of AI apps granted OAuth access to Google Workspace or Microsoft 365, or revoking OAuth grants, on the shadow AI discovery page.Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-33** (none-found): None found on 2026-11-16: a per-tool data-use or training-on-inputs risk rating, an exportable inventory report, or network, SaaS API, email or billing vantage points, on the shadow AI discovery page.Source: [bay.io](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-34** (observed): Bay documentation at docs.bay.io redirects to a sign-in page and is open only to signed-in Bay users.Quote: "The Bay documentation is open only to signed-in Bay users."Source: [docs.bay.io](https://docs.bay.io/?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-35** (none-found): None found on 2026-11-16: a published price. The site has no pricing page (bay.io/pricing returns 404) and the demo page only offers a 30-minute demo.Source: [bay.io](https://www.bay.io/demo?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-36** (none-found): None found on 2026-11-16: published research, threat reports or open-source code; the Bay blog lists only two announcement posts dated 2026-11-10.Source: [bay.io](https://www.bay.io/blog?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-69** (vendor): Bay's product walkthrough names a Claude Code install as a discovered finding.Quote: "A Claude Code install is signed in with a personal Gmail account and connected to Salesforce, Atlassian and Sentry."Source: [bay.io](https://www.bay.io/platform?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-70** (vendor): Bay names Claude Code and Cursor as coding agents it secures on the endpoint.Quote: "How do you secure AI coding agents like Claude Code and Cursor on the endpoint?"Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. - **bay-72** (vendor): Bay describes its endpoint scope as coding agents, desktop assistants, MCP servers and skills on employee devices.Quote: "Coding agents, desktop assistants, MCP servers and skills, working with access to code, data and production environments."Source: [bay.io](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=vendor), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Aurascape profile](/vendors/aurascape/) - [Cyberhaven profile](/vendors/cyberhaven/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Cyberhaven for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/cyberhaven/ Updated: 2026-11-16 Cyberhaven finds shadow AI with an endpoint agent and a browser extension, and adds API connectors for ChatGPT Enterprise and Claude Enterprise. Its product pages name SaaS GenAI apps, IDE and command-line coding assistants, agent frameworks and MCP servers, scored on five risk dimensions, with who uses each tool and with what data. Documented actions are block, warn, redact, coach and redirect, with ServiceNow and SOAR integrations. A free public AI risk checker is published. None found on 16 Nov 2026 for discovery of browser extensions, OAuth-granted AI apps or AI inside approved SaaS, for OAuth revocation, or for a published price, public docs or time to a first inventory. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Endpoint-agent inventory of AI apps, coding assistants, agent frameworks and MCP servers with a five-dimension risk score and block, warn, redact and coach actions (cyberhaven-01, cyberhaven-02, cyberhaven-21, cyberhaven-16, cyberhaven-17). **Weakness:** None found on 2026-11-16 for AI browser extensions, OAuth-granted AI apps or AI inside approved SaaS (cyberhaven-07), a network or proxy vantage (cyberhaven-12), a time to first inventory (cyberhaven-14), OAuth revocation (cyberhaven-20), a published price (cyberhaven-26) or public docs (cyberhaven-27). ## Scores and evidence Cyberhaven is 8th in the overall ranking at 54.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Cyberhaven: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | Endpoint discovery of SaaS genAI apps, IDE and CLI coding assistants, agent frameworks and MCP servers, plus enterprise AI compliance APIs; Help Net Security confirms endpoint and browser discovery. None found for browser extensions, OAuth apps or embedded SaaS AI. | [cyberhaven-01](#fact-cyberhaven-01), [cyberhaven-05](#fact-cyberhaven-05), [cyberhaven-03](#fact-cyberhaven-03), [cyberhaven-08](#fact-cyberhaven-08), [cyberhaven-36](#fact-cyberhaven-36), [cyberhaven-07](#fact-cyberhaven-07) | | Where it looks | 5 of 10 | 20 | Endpoint agent, browser extension and named compliance API connectors, with IdP used for context only. None found for network, email or billing sources. | [cyberhaven-01](#fact-cyberhaven-01), [cyberhaven-08](#fact-cyberhaven-08), [cyberhaven-09](#fact-cyberhaven-09), [cyberhaven-13](#fact-cyberhaven-13), [cyberhaven-10](#fact-cyberhaven-10), [cyberhaven-12](#fact-cyberhaven-12) | | Time to a first inventory | 4 of 10 | 15 | Vendor says flows show up without configuration, but an endpoint agent is needed. None found for a time to first inventory or an OS list. | [cyberhaven-11](#fact-cyberhaven-11), [cyberhaven-13](#fact-cyberhaven-13), [cyberhaven-14](#fact-cyberhaven-14), [cyberhaven-15](#fact-cyberhaven-15) | | What it can do about a finding | 7 of 10 | 15 | Block, warn, redact, coach and redirect, plus ServiceNow and SOAR integrations, all vendor documented. None found for OAuth revocation or an approval workflow. | [cyberhaven-16](#fact-cyberhaven-16), [cyberhaven-17](#fact-cyberhaven-17), [cyberhaven-18](#fact-cyberhaven-18), [cyberhaven-19](#fact-cyberhaven-19), [cyberhaven-20](#fact-cyberhaven-20) | | Risk context and reporting | 7 of 10 | 10 | Per-app risk scores on five dimensions, sanction categories, who uses each tool and with what data, and audit records; vendor claims only. | [cyberhaven-21](#fact-cyberhaven-21), [cyberhaven-22](#fact-cyberhaven-22), [cyberhaven-23](#fact-cyberhaven-23), [cyberhaven-24](#fact-cyberhaven-24), [cyberhaven-25](#fact-cyberhaven-25) | | Transparency | 3 of 10 | 10 | Published research and a free public risk checker. None found for a price or public documentation. | [cyberhaven-25](#fact-cyberhaven-25), [cyberhaven-28](#fact-cyberhaven-28), [cyberhaven-29](#fact-cyberhaven-29), [cyberhaven-26](#fact-cyberhaven-26), [cyberhaven-27](#fact-cyberhaven-27) | | Proof | 6 of 10 | 5 | Named customers on the vendor page and a customer quoted by VentureBeat; analyst mention only in a vendor release. None found for funding figures. | [cyberhaven-30](#fact-cyberhaven-30), [cyberhaven-32](#fact-cyberhaven-32), [cyberhaven-33](#fact-cyberhaven-33), [cyberhaven-34](#fact-cyberhaven-34), [cyberhaven-35](#fact-cyberhaven-35) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 8th | 54.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 6th | 59.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 10th | 51.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 7th | 56.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | joint 7th | 56.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 5th | 59.5 | ## Product, deployment, price and owner - Product assessed: Cyberhaven Flow: Agentic Data Security (AI security: shadow AI discovery, AI Risk IQ scoring, AI data flow control), with the endpoint agent and browser extension - Deployment: Endpoint agent (sensor) for discovery and enforcement; standalone browser extension for ChromeOS and unmanaged devices; API cloud connectors for ChatGPT Enterprise and Claude Enterprise - Time to deploy: none found on 2026-11-16 - Price: none found on 2026-11-16 (pricing page returns 404; demo request only; sold also via AWS, Azure and Google Cloud marketplaces) - Owner: Cyberhaven (standalone company, Mountain View, CA; no parent named on its pages) - Website: cyberhaven.com ## Facts and sources - **cyberhaven-01** (vendor): Cyberhaven says its endpoint agent continuously inventories AI tools, covering SaaS generative AI apps, coding assistants in IDEs and CLIs, open-source agent frameworks and MCP servers.Quote: "continuously inventory AI tools as they appear across the organization, including mainstream SaaS generative AI applications, coding assistants running in IDEs and CLIs"Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-02** (vendor): Cyberhaven says it discovers and monitors MCP servers and AI connectors across the enterprise.Quote: "Discovers and monitors Model Context Protocol servers and AI connectors across the enterprise, surfacing risk from integrations that operate outside traditional security controls."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-03** (vendor): Cyberhaven says its Agentic AI Security (announced 2026-05-05) keeps an inventory of AI agents, GenAI apps and MCP servers, including shadow agents on endpoints, with Risk IQ scores.Quote: "provides a continuously maintained inventory of new and emerging AI agents, GenAI applications, and MCP servers across the enterprise, including shadow agents running locally on endpoints"Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-solves-the-surge-in-shadow-ai), read 2026-11-16. - **cyberhaven-04** (vendor): Cyberhaven says shadow agents run locally in developer tools, IDEs and desktop applications, which is where its endpoint discovery looks.Quote: "Most security tools built for cloud-based AI are blind to agents running locally on endpoints in developer tools, integrated development environments (IDEs), and desktop applications."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-solves-the-surge-in-shadow-ai), read 2026-11-16. - **cyberhaven-05** (vendor): Cyberhaven says its endpoint agent monitors coding assistants such as Claude Code and Copilot.Quote: "Cyberhaven's endpoint agent monitors agentic AI tools including coding assistants like Claude Code and Copilot"Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-06** (vendor): On 2026-11-05, Cyberhaven launched Agentic Data Security and said it discovers AI tools and agents across endpoints, any cloud or browser, including unsanctioned apps, and scores each for risk.Quote: "Discover every AI tool and agent running across endpoints, any cloud or browser, including unsanctioned apps, and score each one for risk."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-delivers-agentic-data-security), read 2026-11-16. - **cyberhaven-07** (none-found): None found on 2026-11-16: documented discovery of AI browser extensions, AI browsers, AI apps granted OAuth access, or AI features embedded in approved SaaS (such as Notion AI or Slack AI) on the AI security product page, the secure-AI use case page or the generative AI page.Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-08** (vendor): On 2026-07-23, Cyberhaven said it reads ChatGPT Enterprise and Claude Enterprise through their Compliance APIs, covering conversations, uploaded files and custom GPT configurations already inside each platform.Quote: "The integrations use the ChatGPT Enterprise Compliance API and Claude Compliance APIs to monitor, inspect, and classify sensitive data already inside each platform."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-extends-ai-security-chatgpt-enterprise-claude), read 2026-11-16. - **cyberhaven-09** (vendor): Cyberhaven says it connects to ChatGPT Enterprise and Claude Enterprise directly and inspects every other AI tool at the endpoint and in the browser.Quote: "Cyberhaven connects to ChatGPT Enterprise and Claude Enterprise directly, and inspects every other AI tool at the endpoint and in the browser."Source: [cyberhaven.com](https://www.cyberhaven.com/product/integrations), read 2026-11-16. - **cyberhaven-10** (vendor): Cyberhaven says it pulls directory context (departments, roles, groups) from identity providers such as Okta, Entra ID and Google Workspace for policies and incident review; this is context, not documented discovery.Quote: "Build policies around the departments, roles, and groups you already maintain in your directory, and see that context when reviewing incidents."Source: [cyberhaven.com](https://www.cyberhaven.com/product/integrations), read 2026-11-16. - **cyberhaven-11** (vendor): Cyberhaven says it records all data flows to the internet without configuration or policies, so AI tool use shows up out of the box.Quote: "Cyberhaven records all data flows to the internet without any configuration or policies needed, so your team can understand what sensitive data is flowing to and from these tools."Source: [cyberhaven.com](https://www.cyberhaven.com/technologies/gen-ai), read 2026-11-16. - **cyberhaven-12** (none-found): None found on 2026-11-16: network, proxy, firewall or DNS traffic as a discovery vantage point, or email and billing records, on the AI security product page; the page says network-layer tools cannot see OS-level agents and positions the endpoint agent as the sensor.Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-13** (vendor): Cyberhaven says a standalone browser extension (announced 2026-05-05) extends its data loss prevention to ChromeOS, contractor devices and unmanaged endpoints with no endpoint sensor required.Quote: "Extends data loss prevention to ChromeOS, contractor devices, and unmanaged endpoints with no endpoint sensor required."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-solves-the-surge-in-shadow-ai), read 2026-11-16. - **cyberhaven-14** (none-found): None found on 2026-11-16: a published time to first AI inventory after setup on the AI security product page, the generative AI page or the press releases read.Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-15** (none-found): None found on 2026-11-16: a list of supported operating systems or browsers for the endpoint agent on the AI security product page (ChromeOS is named only for the standalone browser extension).Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-16** (vendor): Cyberhaven says it enforces guardrails at the prompt and response level: blocking high-risk data movement, redirecting users to sanctioned tools and coaching with plain-English explanations.Quote: "Enforces runtime guardrails at the prompt and response level, blocking high-risk data movement, redirecting users to sanctioned tools, and coaching employees with plain-English policy explanations."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-17** (vendor): Cyberhaven says it offers block, warn and redact controls for risky agent actions across the agentic workflow.Quote: "Stops unsafe or high-risk agent actions before they cause damage, with block, warn, and redact controls that operate across the full agentic workflow."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-18** (vendor): Cyberhaven says that when data is headed to an unapproved AI app it can show the user a policy message and redirect them to approved alternatives.Quote: "When data is at risk of flowing to an unapproved AI application, instantly take action and surface a message to the user educating them on company policy and redirect them to approved alternatives."Source: [cyberhaven.com](https://www.cyberhaven.com/technologies/gen-ai), read 2026-11-16. - **cyberhaven-19** (vendor): Cyberhaven says it streams incidents, events and audit logs to SIEM, SOAR and AI SOC platforms, with purpose-built Splunk, Cortex XSOAR and ServiceNow integrations.Quote: "Stream incidents, events, and audit logs into your SIEM, SOAR, or AI SOC platform. Splunk, Cortex XSOAR, and ServiceNow have purpose-built integrations"Source: [cyberhaven.com](https://www.cyberhaven.com/product/integrations), read 2026-11-16. - **cyberhaven-20** (none-found): None found on 2026-11-16: revoking OAuth grants to AI apps, or an approval workflow that requires sign-off before an AI tool is used, on the AI security product page or the secure-AI use case page.Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-21** (vendor): Cyberhaven says its AI Risk IQ scores each AI app and agent on five dimensions, maintained quarterly by its research team.Quote: "Cyberhaven's AI Risk IQ scores every AI application and agent across five dimensions: data sensitivity, model integrity, compliance adherence, user access controls, and security infrastructure."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-22** (vendor): Cyberhaven says it categorizes AI applications as Sanctioned, Unsanctioned, Tolerated or Restricted and shows adoption trends.Quote: "Surfaces AI adoption trends across the enterprise, categorizing applications as Sanctioned, Unsanctioned, Tolerated, or Restricted to support governance decisions."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-23** (vendor): Cyberhaven says it shows every AI tool used, who is using it and what kinds of data interact with it.Quote: "Cyberhaven lets you see every AI tool used by your organization, who’s using them, and what kinds of data are interacting with them."Source: [cyberhaven.com](https://www.cyberhaven.com/technologies/gen-ai), read 2026-11-16. - **cyberhaven-24** (vendor): Cyberhaven says it keeps a record of AI agent activity, data access and policy enforcement for audit and regulatory reporting.Quote: "Provides a comprehensive record of AI agent activity, data access, and policy enforcement actions to support audit workflows and regulatory reporting."Source: [cyberhaven.com](https://www.cyberhaven.com/product/ai-security), read 2026-11-16. - **cyberhaven-25** (vendor): Cyberhaven offers a free public AI Risk Checker with risk profiles for 700+ AI apps scored Very Low to Critical, with 3 free checks.Quote: "Risk profiles for 700+ AI applications"Source: [cyberhaven.com](https://www.cyberhaven.com/labs/ai-risk-checker), read 2026-11-16. - **cyberhaven-26** (none-found): None found on 2026-11-16: a published price or pricing page; https://www.cyberhaven.com/pricing returned 404 Not Found and pages offer only Request demo.Source: [cyberhaven.com](https://www.cyberhaven.com/pricing), read 2026-11-16. - **cyberhaven-27** (none-found): None found on 2026-11-16: public product documentation readable without sign-in; docs.cyberhaven.com did not load, the documentation link points to a sign-in host (docs-auth.cyberhaven.io) and the support site returned no readable text.Source: [support.cyberhaven.com](https://support.cyberhaven.com/), read 2026-11-16. - **cyberhaven-28** (third-party): In a newswire release, Cyberhaven Labs reported an analysis of 3 million workers finding AI usage grew 485% year over year, with over 90% of use in personal shadow AI accounts (a 2024 report, undated here).Quote: "Analysis covering 3 million workers finds AI usage grew 485% year-on-year with over 90% of usage in personal "shadow AI" accounts"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/cyberhaven-report-surge-in-shadow-ai-accounts-poses-fresh-risks-to-corporate-data-302151221.html), read 2026-11-16. - **cyberhaven-29** (vendor): Cyberhaven says Cyberhaven Labs research found endpoint-based AI-native app adoption grew 509% and coding assistant adoption grew 357% year over year (release dated 2026-05-05).Quote: "enterprise adoption of endpoint-based AI-native apps has grown 509% over the past year"Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-solves-the-surge-in-shadow-ai), read 2026-11-16. - **cyberhaven-30** (vendor): Cyberhaven's customers page names BioIVT, Axos Bank, Plaid, Cooley and Mission Australia in customer quotes and stories.Quote: "BioIVT Building confidence in data lineage and AI security across a global life sciences organization."Source: [cyberhaven.com](https://www.cyberhaven.com/customers), read 2026-11-16. - **cyberhaven-31** (vendor): Plaid's Head of Platform Security is quoted on Cyberhaven's customers page about visibility gained.Quote: "Cyberhaven created a transformational difference. We now have really substantial visibility."Source: [cyberhaven.com](https://www.cyberhaven.com/customers), read 2026-11-16. - **cyberhaven-32** (third-party): VentureBeat quoted Greenlight's CSO and CIO on Cyberhaven in its coverage of the Linea AI launch (article date not read).Quote: "Cyberhaven shows us exactly how our data moves and is used across the organization, giving us visibility not found with traditional security tools"Source: [venturebeat.com](https://venturebeat.com/security/security-teams-can-respond-80-faster-to-events-with-cyberhavens-ai-powered-data-lineage-tools), read 2026-11-16. - **cyberhaven-33** (vendor): A November 2026 Cyberhaven release quotes Lawrence Pingree of analyst firm SACR on its data lineage; this is an analyst quote in a vendor release, not an independent report.Quote: "Cyberhaven’s data lineage gives security teams that foundation for people and AI agents together, in one platform."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-delivers-agentic-data-security), read 2026-11-16. - **cyberhaven-34** (vendor): Cyberhaven's newsroom features a Forbes headline calling it a "$1 Billion" startup and lists a Fortune Cyber 60 award; no funding round figures were read on any page fetched.Quote: "This $1 Billion Cyber Startup Stops Employees From Leaking Vital Data To AI"Source: [cyberhaven.com](https://www.cyberhaven.com/newsroom), read 2026-11-16. - **cyberhaven-35** (vendor): On 2026-04-22, Cyberhaven said it became available on AWS, Azure and Google Cloud marketplaces.Quote: "Cyberhaven Now Available on AWS, Azure, and Google Cloud Marketplaces, Giving Enterprises Flexible Procurement for AI-Native Data Security"Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-now-available-aws-azure-google-cloud-marketplaces), read 2026-11-16. - **cyberhaven-36** (third-party): On 2026-07-28, Help Net Security reported the Flow launch, saying Cyberhaven discovers AI tools and agents on endpoints and in the browser and scores each for risk.Quote: "Cyberhaven discovers every AI tool and agent running across endpoints and the browser, scores each one for risk"Source: [helpnetsecurity.com](https://www.helpnetsecurity.com/2026/07/28/cyberhaven-flow/), read 2026-11-16. - **cyberhaven-37** (vendor): On 2026-07-28, Cyberhaven introduced Flow, an AI-native data security platform; Mountain View, CA is the dateline and the release said Flow would be available in the coming quarter.The exact quote is on the source page.Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-introduces-flow-ai-native-data-security), read 2026-11-16. - **cyberhaven-38** (vendor): On 2026-11-05, Cyberhaven said Agentic Data Security is the latest module in Cyberhaven Flow and is available for purchase today.Quote: "Agentic Data Security is the latest module in Cyberhaven Flow"Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-delivers-agentic-data-security), read 2026-11-16. - **cyberhaven-39** (vendor): On 2026-08-19, Cyberhaven named Todd Rotger President, Go-to-Market and added Manish Kalia as Chief Product Officer; its about page lists Nishant Doshi as CEO.Quote: "today announced the appointment of Todd Rotger as President, Go-to-Market."Source: [cyberhaven.com](https://www.cyberhaven.com/press-releases/cyberhaven-names-todd-rotger-president-to-lead-next-phase-of-growth), read 2026-11-16. - **cyberhaven-40** (vendor): Cyberhaven's own about page describes the company as protecting data across human and agentic workflows and lists no parent company.Quote: "Cyberhaven protects data across human and agentic workflows, adapting protection to changing context."Source: [cyberhaven.com](https://www.cyberhaven.com/about), read 2026-11-16. ## News mentioning Cyberhaven - [Cyberhaven launches Agentic Data Security module](/news/cyberhaven-launches-agentic-data-security/) ## Related pages - [All vendor profiles](/vendors/) - [Bay profile](/vendors/bay/) - [Grip Security profile](/vendors/grip-security/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Grip Security for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/grip-security/ Updated: 2026-11-16 Grip Security starts from identity. It connects to email, the identity provider and SaaS applications through read-only APIs, with an optional browser extension called Grip Extend, and says setup takes 10 minutes with about 5 days to gather historic usage. Its documentation says it finds AI applications, agents, embedded AI in approved SaaS and OAuth-scoped apps, and it lists connectors for Custom GPTs in ChatGPT Enterprise and Gemini agents. Grip says it can categorize OAuth scopes by risk and revoke access, and it offers a free Shadow AI Analysis. None found on 16 Nov 2026 for endpoint, MCP server or coding agent discovery. Pricing is per human user per year, with no amounts published. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Identity-driven discovery of AI apps, agents, embedded SaaS AI and risky OAuth access with a 10-minute setup and a free Shadow AI Analysis (grip-security-05, grip-security-06, grip-security-14). **Weakness:** None found on 2026-11-16 for endpoint, MCP server or coding agent discovery, or in-the-moment blocking of data sent to AI tools (grip-security-09, grip-security-20). ## Scores and evidence Grip Security is 5th in the overall ranking at 58.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Grip Security: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 5 of 10 | 25 | AI apps, agents, extensions, embedded AI in approved SaaS, OAuth-scoped apps and custom GPTs are claimed. None found for MCP servers, desktop AI apps, IDE plugins or CLI agents. | [grip-security-05](#fact-grip-security-05), [grip-security-06](#fact-grip-security-06), [grip-security-07](#fact-grip-security-07), [grip-security-08](#fact-grip-security-08), [grip-security-09](#fact-grip-security-09) | | Where it looks | 7 of 10 | 20 | Email, identity provider, SaaS APIs and an optional browser extension, with named integrations such as Okta, Entra ID, Google Workspace and ChatGPT Enterprise; vendor claims only. | [grip-security-10](#fact-grip-security-10), [grip-security-11](#fact-grip-security-11), [grip-security-12](#fact-grip-security-12), [grip-security-13](#fact-grip-security-13) | | Time to a first inventory | 7 of 10 | 15 | Vendor publishes 10-minute agentless setup via read-only APIs and about 5 days for historic usage. None found for an OS or browser list. | [grip-security-14](#fact-grip-security-14), [grip-security-15](#fact-grip-security-15), [grip-security-12](#fact-grip-security-12), [grip-security-16](#fact-grip-security-16) | | What it can do about a finding | 6 of 10 | 15 | OAuth revocation, approval workflows, user prompts, blocking via a named gateway integration and Jira tickets. None found for in-the-moment prompt blocking or redaction. | [grip-security-17](#fact-grip-security-17), [grip-security-18](#fact-grip-security-18), [grip-security-19](#fact-grip-security-19), [grip-security-07](#fact-grip-security-07), [grip-security-20](#fact-grip-security-20) | | Risk context and reporting | 4 of 10 | 10 | Identity-based risk scoring and an inventory mapped to identities. None found for a data-use or training rating or an export format. | [grip-security-21](#fact-grip-security-21), [grip-security-22](#fact-grip-security-22), [grip-security-23](#fact-grip-security-23) | | Transparency | 5 of 10 | 10 | Public use-case and integration pages, a pricing page without amounts and published risk research, with the newest report gated. | [grip-security-24](#fact-grip-security-24), [grip-security-25](#fact-grip-security-25), [grip-security-26](#fact-grip-security-26) | | Proof | 7 of 10 | 5 | Series B and $66 million total reported by Help Net Security, CRN coverage and named customers. None found for independent analyst reports. | [grip-security-28](#fact-grip-security-28), [grip-security-29](#fact-grip-security-29), [grip-security-30](#fact-grip-security-30), [grip-security-27](#fact-grip-security-27), [grip-security-32](#fact-grip-security-32) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 5th | 58.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 8th | 55.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 5th | 61.0 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 6th | 59.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 5th | 59.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 6th | 58.5 | ## Product, deployment, price and owner - Product assessed: Grip AI + SaaS Security Control Plane: Shadow AI Discovery, free Shadow AI Analysis, Grip Extend - Deployment: Agentless API connections to email, identity provider and SaaS; optional browser extension - Time to deploy: 10 minutes to deploy; about 5 days to gather historic usage - Price: Per human user, annual, custom pricing; no amounts published - Owner: Grip Security, Inc. - Website: [grip.security](https://www.grip.security) ## Facts and sources - **grip-security-01** (vendor): The site is operated by Grip Security, Inc., and no parent company or acquirer is named on its pages.Quote: "Copyright (C) 2026 Grip Security, Inc."Source: [grip.security](https://www.grip.security), read 2026-11-16. - **grip-security-02** (vendor): Grip says it was founded to bring innovation to SaaS security, and lists Lior Yaari as CEO and co-founder and Idan Fast as CTO and co-founder.Quote: "Grip was founded to bring much-needed innovation to SaaS security."Source: [grip.security](https://www.grip.security/saas-security-about-us), read 2026-11-16. - **grip-security-03** (vendor): Grip lists offices in Boston, Massachusetts and Tel Aviv, Israel, and holds SOC 2 Type II and ISO 27001 certifications.Quote: "50 Milk Street, 16th Floor Boston, MA 02109"Source: [grip.security](https://www.grip.security), read 2026-11-16. - **grip-security-04** (vendor): Grip positions its product as an AI and SaaS security control plane that builds an app inventory and links assets to identities.Quote: "Grip builds an app inventory and links assets to identities. It is the technical backbone for AI governance"Source: [grip.security](https://www.grip.security), read 2026-11-16. - **grip-security-05** (vendor): Grip says its shadow AI discovery identifies AI applications, agents, extensions, embedded capabilities and accounts across the enterprise.Quote: "Identify AI applications, agents, extensions, embedded capabilities, and accounts in use across the enterprise."Source: [grip.security](https://www.grip.security/use-case-library/discover-and-control-shadow-ai-across-your-enterprise), read 2026-11-16. - **grip-security-06** (vendor): Grip says it flags existing approved SaaS that has AI features and detects unauthorized AI apps.Quote: "detect unauthorized AI apps and flag existing, approved SaaS with AI features."Source: [grip.security](https://www.grip.security/use-case-library/reduce-shadow-ai-risks), read 2026-11-16. - **grip-security-07** (vendor): Grip says it categorizes OAuth scopes by risk and can revoke access entirely, which covers AI apps granted OAuth access among other apps.Quote: "Grip categorizes OAuth scopes and permissions by risk level, allowing you to prioritize and address the most critical threats, including revoking access entirely."Source: [grip.security](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16. - **grip-security-08** (vendor): On its integrations page Grip lists connectors that discover Custom GPTs in ChatGPT Enterprise and Gemini agents in Google Gemini.Quote: "Discovers Custom GPTs, tracks ownership and AI identities, maps connectivity, and manages posture risk."Source: [grip.security](https://www.grip.security/integrations), read 2026-11-16. - **grip-security-09** (none-found): None found on 2026-11-16: Grip's pages name no coverage of MCP servers on endpoints, desktop AI apps, or IDE plugins and command-line coding agents; its MCP page covers only GripMCP, its own interface for querying Grip data with AI.Source: [grip.security](https://www.grip.security/mcp-security), read 2026-11-16. - **grip-security-10** (vendor): Grip says discovery connects the customer's email and identity provider, with an optional browser extension called Grip Extend for deeper insight.Quote: "Connect your email and identity provider in just 10 minutes to uncover past, current, and future apps. Add the optional Grip Extend browser extension for deeper insights."Source: [grip.security](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16. - **grip-security-11** (vendor): Grip's integrations page lists identity, SaaS and security connectors including Okta, Entra ID, Google Workspace, Claude, ChatGPT Enterprise, Google Gemini, CrowdStrike and Cisco Umbrella.Quote: "Block access to risky or unapproved apps discovered by Grip."Source: [grip.security](https://www.grip.security/integrations), read 2026-11-16. - **grip-security-12** (vendor): Grip says its Okta integration deploys in 10 minutes with no agents or proxies.Quote: "10 minute deployment with no agents or proxies"Source: [grip.security](https://www.grip.security/okta-integration), read 2026-11-16. - **grip-security-13** (vendor): Grip describes its browser extension as detecting credential risks, shared accounts and apps without MFA, and as engaging users before risky actions happen.Quote: "Engage with users and prevent risky actions before they happen."Source: [grip.security](https://www.grip.security/platform/browser-extension), read 2026-11-16. - **grip-security-14** (vendor): Grip says it deploys in 10 minutes and usually needs about 5 days to gather historic and current AI and SaaS usage during a proof of concept.Quote: "Grip deploys in 10 minutes. Depending on your organization’s size, it will likely take 5 days to fully gather all historic and current AI + SaaS usage."Source: [grip.security](https://www.grip.security/ai-security), read 2026-11-16. - **grip-security-15** (vendor): Grip says its free Shadow AI Analysis sets up in about 10 minutes through read-only APIs and the point-in-time report takes 5 to 10 days.The exact quote is on the source page.Source: [grip.security](https://www.grip.security/pricing), read 2026-11-16. - **grip-security-16** (none-found): None found on 2026-11-16: Grip's public pages list no supported operating systems or browsers for Grip Extend, and name no Chrome Web Store or other marketplace listing in the pages read.Source: [grip.security](https://www.grip.security/platform/browser-extension), read 2026-11-16. - **grip-security-17** (vendor): Grip says it can apply SSO, MFA or revoke risky access using automated password rotation, and automate mitigation through built-in workflows.Quote: "Apply controls like SSO, MFA, or revoke risky access using automated password rotation"Source: [grip.security](https://www.grip.security/use-case-library/reduce-shadow-ai-risks), read 2026-11-16. - **grip-security-18** (vendor): Grip lists workflow actions including notifying on unsanctioned app signups, approval workflows for onboarding, targeted user prompts and Slack alerts.Quote: "Trigger approval workflows for app onboarding and access governance"Source: [grip.security](https://www.grip.security/ai-security), read 2026-11-16. - **grip-security-19** (vendor): Grip lists a Cisco Umbrella integration that blocks access to risky or unapproved apps Grip discovers, and a Jira integration that creates tickets.Quote: "Integrate shadow SaaS management into Jira processes by automating ticket creation as needed."Source: [grip.security](https://www.grip.security/integrations), read 2026-11-16. - **grip-security-20** (none-found): None found on 2026-11-16: Grip's pages read name no in-the-moment blocking of prompts or redaction of data sent to AI tools; its data exposure page frames controls as approve, restrict, investigate or remediate.Source: [grip.security](https://www.grip.security/use-case-library/prevent-sensitive-data-exposure-to-ai), read 2026-11-16. - **grip-security-21** (vendor): Grip says it assesses each AI app's risk from authentication methods, usage patterns and other identity risk factors, with risk scoring and recommended actions.Quote: "Assess each app’s risk based on authentication methods, usage patterns, and other identity risk factors to understand the risks AI apps pose."Source: [grip.security](https://www.grip.security/use-case-library/reduce-shadow-ai-risks), read 2026-11-16. - **grip-security-22** (vendor): Grip says its free Shadow AI Analysis delivers an inventory of discovered AI apps, agents, extensions and integrations mapped to identities and access.Quote: "Inventory discovered AI apps, agents, extensions, and integrations"Source: [grip.security](https://www.grip.security/pricing), read 2026-11-16. - **grip-security-23** (none-found): None found on 2026-11-16: no per-tool rating of data use or training on inputs, and no description of an exportable inventory file format, on the Grip pages read.Source: [grip.security](https://www.grip.security/ai-security), read 2026-11-16. - **grip-security-24** (vendor): Grip publishes a pricing page stating per-human annual pricing, but gives no dollar amounts and routes SMB and enterprise tiers to a demo.Quote: "Grip is charged on an annual, per user basis."Source: [grip.security](https://www.grip.security/pricing), read 2026-11-16. - **grip-security-25** (vendor): Grip publishes use-case pages, a glossary and integration pages readable without sign-in, but gates its 2026 report behind a form.Quote: "Get the Free Report"Source: [grip.security](https://www.grip.security/saas-ai-governance-report-2026), read 2026-11-16. - **grip-security-26** (vendor): Grip says its 2025 risk report found 90 percent of SaaS applications and 91 percent of AI tools unmanaged (press release dated October 23, 2024).Quote: "Grip Security Releases 2025 SaaS Security Risks Report: Reveals 90% of SaaS Applications and 91% of AI Tools are Unmanaged"Source: [grip.security](https://www.grip.security/press-releases-lobby), read 2026-11-16. - **grip-security-27** (vendor): Grip says it is scored 4.9 out of 5 on Gartner and protects 125 million users.The exact quote is on the source page.Source: [grip.security](https://www.grip.security), read 2026-11-16. - **grip-security-28** (third-party): Help Net Security reported that Grip raised a $41 million Series B led by Third Point Ventures, bringing total funding to $66 million.Quote: "Grip Security raised $41 million in Series B funding led by Third Point Ventures"Source: [helpnetsecurity.com](https://www.helpnetsecurity.com/2023/08/22/grip-security-funding/), read 2026-11-16. - **grip-security-29** (third-party): CRN included Grip in a Black Hat 2024 roundup noting its partnership with GuidePoint Security.Quote: "Grip Security, an identity risk management startup focused on SaaS, also announced a new channel move"Source: [crn.com](https://www.crn.com/news/security/2024/10-security-vendors-making-moves-at-black-hat-2024?page=8), read 2026-11-16. - **grip-security-30** (vendor): Grip's success stories page names Believer, PDS Health, Interpublic Group and Texas Baptists among customers.Quote: "Believer Secured Innovation by Containing Shadow AI Risks"Source: [grip.security](https://www.grip.security/success-story), read 2026-11-16. - **grip-security-31** (none-found): None found on 2026-11-16: no news, launch or funding item dated 2026-08-18 or later on Grip's press and news pages; the newest press release listed is dated June 5, 2025.Source: [grip.security](https://www.grip.security/press-releases-lobby), read 2026-11-16. - **grip-security-32** (third-party): Help Net Security reported Grip's SSPM launch on 2025-02-14, a posture management product built on visibility and automation.Quote: "Grip Security has unveiled its SaaS Security Posture Management (SSPM) solution"Source: [helpnetsecurity.com](https://www.helpnetsecurity.com/2025/02/14/grip-security-unveils-sspm-solution-to-strengthen-saas-security-posture/), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Cyberhaven profile](/vendors/cyberhaven/) - [Harmonic Security profile](/vendors/harmonic-security/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Harmonic Security for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/harmonic-security/ Updated: 2026-11-16 Harmonic Security sells Harmonic Explore for discovery and risk scoring and Harmonic Guide for inline coaching, fed by a browser extension and an endpoint agent pushed through Intune, Jamf or Kandji. Harmonic says first insight arrives in about 30 minutes, and its documentation lists Windows, macOS and Linux. Its endpoint page names desktop apps, coding assistants, command-line tools, local models and MCP servers, and its docs describe a per-app risk profile with CSV export. Actions include warn, nudge and block, personal-account restriction and webhooks for ticketing. None found on 16 Nov 2026 for discovery of apps granted OAuth access, email or billing records, or a published price: the pricing page lists Explore, Guide and Command tiers without amounts. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Broad discovery of web, desktop, coding and MCP AI tools with a documented per-app risk score and a stated first insight in about 30 minutes (harmonic-security-10, harmonic-security-12, harmonic-security-24, harmonic-security-33). **Weakness:** None found on 2026-11-16 for discovering AI apps granted OAuth access or for email and billing records, and no price is published (harmonic-security-16, harmonic-security-23, harmonic-security-38). ## Scores and evidence Harmonic Security is 1st in the overall ranking at 73.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Harmonic Security: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 7 of 10 | 25 | Desktop apps, coding assistants, CLI tools, local models, named MCP clients and servers, AI browsers and embedded SaaS AI; docs observe desktop AI reporting. None found for OAuth-granted AI apps. | [harmonic-security-12](#fact-harmonic-security-12), [harmonic-security-14](#fact-harmonic-security-14), [harmonic-security-15](#fact-harmonic-security-15), [harmonic-security-10](#fact-harmonic-security-10), [harmonic-security-22](#fact-harmonic-security-22), [harmonic-security-11](#fact-harmonic-security-11), [harmonic-security-16](#fact-harmonic-security-16) | | Where it looks | 7 of 10 | 20 | Endpoint agent, browser extension, AI network hosts and an observed compliance API connector; the IdP only enriches. None found for email or billing records. | [harmonic-security-20](#fact-harmonic-security-20), [harmonic-security-21](#fact-harmonic-security-21), [harmonic-security-19](#fact-harmonic-security-19), [harmonic-security-22](#fact-harmonic-security-22), [harmonic-security-18](#fact-harmonic-security-18), [harmonic-security-23](#fact-harmonic-security-23) | | Time to a first inventory | 8 of 10 | 15 | Vendor publishes first insight in about 30 minutes via MDM, docs observe an OS list, and browsers are named; a full map takes days. | [harmonic-security-24](#fact-harmonic-security-24), [harmonic-security-25](#fact-harmonic-security-25), [harmonic-security-26](#fact-harmonic-security-26), [harmonic-security-15](#fact-harmonic-security-15), [harmonic-security-27](#fact-harmonic-security-27) | | What it can do about a finding | 8 of 10 | 15 | Inline warn, nudge and block, personal-account restriction, observed MCP access requests and observed webhooks for ticketing. None found for OAuth revocation. | [harmonic-security-28](#fact-harmonic-security-28), [harmonic-security-29](#fact-harmonic-security-29), [harmonic-security-32](#fact-harmonic-security-32), [harmonic-security-30](#fact-harmonic-security-30), [harmonic-security-31](#fact-harmonic-security-31) | | Risk context and reporting | 9 of 10 | 10 | Docs observe per-app risk profiles and CSV export; the vendor adds who uses each tool and whether it trains on inputs. | [harmonic-security-33](#fact-harmonic-security-33), [harmonic-security-34](#fact-harmonic-security-34), [harmonic-security-36](#fact-harmonic-security-36), [harmonic-security-35](#fact-harmonic-security-35) | | Transparency | 6 of 10 | 10 | Documentation observed public and published research. None found for a price. | [harmonic-security-37](#fact-harmonic-security-37), [harmonic-security-39](#fact-harmonic-security-39), [harmonic-security-38](#fact-harmonic-security-38) | | Proof | 6 of 10 | 5 | Funding of $76 million and 100+ companies reported by third parties, and one named customer. None found for analyst coverage. | [harmonic-security-03](#fact-harmonic-security-03), [harmonic-security-04](#fact-harmonic-security-04), [harmonic-security-05](#fact-harmonic-security-05), [harmonic-security-06](#fact-harmonic-security-06), [harmonic-security-09](#fact-harmonic-security-09) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 1st | 73.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 1st | 76.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 1st | 73.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 1st | 72.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 1st | 73.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 1st | 75.5 | ## Product, deployment, price and owner - Product assessed: Harmonic Explore (discovery and risk scoring) and Harmonic Guide (inline coaching), with the Endpoint/Protect Agent and browser extension - Deployment: Browser extension plus Harmonic Protect Agent (endpoint agent) pushed by MDM; local MCP Gateway installed by the agent; OpenTelemetry and API connectors; no network proxy - Time to deploy: Harmonic says first insight in about 30 minutes and deployment in minutes via Intune, JAMF or Kandji - Price: none found on 2026-11-16: pricing page lists Explore, Guide and Command tiers without amounts - Owner: Harmonic Security (standalone; no parent company stated on its pages, which show 2026 Harmonic Security copyright) - Website: harmonic.security ## Facts and sources - **harmonic-security-01** (vendor): Harmonic Security describes itself as selling AI Governance and Control, a control layer for AI use by employees.Quote: "Harmonic Security delivers AI Governance and Control (AIGC), the intelligent control layer that secures and enables the AI-First workforce."Source: [harmonic.security](https://www.harmonic.security/about), read 2026-11-16. - **harmonic-security-02** (vendor): The about page names Alastair Paterson as CEO and co-founder and Bryan Woolgar-O'Neil as CTO and co-founder, and lists no parent company.Quote: "Alastair Paterson CEO & CO-FOUNDER, HARMONIC Bryan Woolgar-O'Neil CTO & CO-FOUNDER, HARMONIC"Source: [harmonic.security](https://www.harmonic.security/about), read 2026-11-16. - **harmonic-security-03** (third-party): On 2026-10-14, SecurityWeek reported a $50 million raise for Harmonic led by Crosspoint Capital Partners, with the founders previously at Digital Shadows.Quote: "The fresh investment round was led by Crosspoint Capital Partners, with additional support from existing investors"Source: [securityweek.com](https://www.securityweek.com/harmonic-security-raises-50-million-for-workforce-ai-protection/), read 2026-11-16. - **harmonic-security-04** (third-party): SecurityBrief reported that Harmonic's Series B brings total funding to $76 million and that more than 100 companies use the platform.Quote: "Harmonic Security has raised USD $50 million in a Series B funding round led by Crosspoint Capital Partners, bringing its total funding to USD $76 million."Source: [securitybrief.co.uk](https://securitybrief.co.uk/story/harmonic-security-raises-usd-50-million-in-series-b), read 2026-11-16. - **harmonic-security-05** (third-party): On 2026-10-14, Unite.AI reported Harmonic's statement that it protects millions of employees at more than 100 companies.Quote: "Harmonic said it protects and enables AI use for millions of employees at more than 100 companies"Source: [unite.ai](https://www.unite.ai/harmonic-security-raises-50m-series-b-to-expand-us-and-uk-teams/), read 2026-11-16. - **harmonic-security-06** (vendor): Harmonic's about page carries a named customer testimonial from Monolithic Power Systems, an IT security director saying it finds unauthorized AI apps.Quote: "Harmonic Security uncovers blindspots, including unauthorized and risky AI applications my employees are using."Source: [harmonic.security](https://www.harmonic.security/about), read 2026-11-16. - **harmonic-security-07** (vendor): Harmonic's about page lists an RSA Innovation Sandbox finalist placement in April 2024 in its company timeline.Quote: "2024 • April RSA Innovation Sandbox Finalists"Source: [harmonic.security](https://www.harmonic.security/about), read 2026-11-16. - **harmonic-security-08** (third-party): SecurityBrief reported that Harmonic is listed among ecosystem integrations for Cisco Cloud Control's Agent Builder through the Model Context Protocol.Quote: "It is listed among ecosystem integrations for Cisco Cloud Control's Agent Builder through Model Context Protocol."Source: [securitybrief.co.uk](https://securitybrief.co.uk/story/harmonic-security-raises-usd-50-million-in-series-b), read 2026-11-16. - **harmonic-security-09** (none-found): None found on 2026-11-16: independent analyst coverage of Harmonic itself (a Gartner or Forrester page naming Harmonic). The Gartner MCP Gateway page on harmonic.security is a title-only vendor repost and does not describe Harmonic.Source: [harmonic.security](https://www.harmonic.security/resources/gartner-innovation-insight-mcp-gateway), read 2026-11-16. - **harmonic-security-10** (vendor): Harmonic Explore gives prompt-level insight across over 1,000 applications, including AI features embedded in SaaS tools.Quote: "View prompt level insights across over 1,000 applications, including embedded AI features inside the SaaS tools your teams already work with."Source: [harmonic.security](https://www.harmonic.security/products/explore), read 2026-11-16. - **harmonic-security-11** (vendor): Harmonic's pricing page says AI tool discovery covers 10,000+ apps, including unsanctioned tools and personal accounts.Quote: "AI tool discovery across 10,000+ apps Including unsanctioned tools and personal accounts"Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-12** (vendor): Harmonic's endpoint page says its Endpoint Agent finds desktop apps, coding assistants, CLI tools, local models and the MCP servers running on developer machines.Quote: "Harmonic covers the full set, from Claude Desktop and ChatGPT Desktop to Cursor, Claude Code, GitHub Copilot, Ollama and the MCP servers running on developer machines."Source: [harmonic.security](https://www.harmonic.security/solutions/endpoint-ai-security), read 2026-11-16. - **harmonic-security-13** (vendor): Harmonic says it continuously monitors for employees using unapproved AI tools or agents.Quote: "Harmonic continually monitors for employees using unapproved AI tools or agents."Source: [harmonic.security](https://www.harmonic.security/solutions/shadow-ai-detection), read 2026-11-16. - **harmonic-security-14** (vendor): Harmonic says it automatically inventories MCP clients (Claude Code, Claude Cowork, Codex, Cursor) and the servers they reach.Quote: "Harmonic automatically discovers and inventories MCP clients such as Claude Code, Claude Cowork, Codex and Cursor, along with the servers they reach"Source: [harmonic.security](https://www.harmonic.security/solutions/ai-agent-security), read 2026-11-16. - **harmonic-security-15** (vendor): Harmonic says its browser extension covers all Chromium-based browsers and Safari, including AI browsers such as ChatGPT Atlas (article dated 2025-10-23).Quote: "It provides unified visibility and control across all Chromium-based browsers as well as Safari."Source: [harmonic.security](https://www.harmonic.security/resources/chatgpt-atlas-protecting-employee-use-of-ai-browsers), read 2026-11-16. - **harmonic-security-16** (none-found): None found on 2026-11-16: a documented way to find AI apps granted OAuth access to Google Workspace or Microsoft 365. Looked on the shadow AI page, the Explore page, and the docs index (llms.txt); only Okta, Entra and Google Workspace integrations that enrich employee data were found.Source: [harmonic.security](https://www.harmonic.security/solutions/shadow-ai-detection), read 2026-11-16. - **harmonic-security-17** (vendor): Harmonic's pricing page lists identity enrichment through Entra, Okta and Google Workspace.Quote: "Identity enrichment through Entra, Okta, and Google Workspace"Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-18** (observed): Harmonic's docs say the Okta integration only enriches employee data (department, location) for filtering and grouping.Quote: "Enrich employee data and enable additional filtering and grouping by integrating Okta."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/llms.txt), read 2026-11-16. - **harmonic-security-19** (observed): Harmonic's docs describe an Anthropic Compliance API connector for prompt-level visibility into Claude Enterprise activity.Quote: "Connect Harmonic to the Anthropic Compliance API to get prompt-level visibility for Claude Enterprise activity across your organization."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/llms.txt), read 2026-11-16. - **harmonic-security-20** (vendor): Harmonic's endpoint page says the Endpoint Agent identifies native AI apps, AI IDEs, CLI tools, local models and direct inference API calls on employee devices.Quote: "identifies every native AI app, AI-first IDE, CLI tool, locally hosted model and direct inference API call running on employee devices"Source: [harmonic.security](https://www.harmonic.security/solutions/endpoint-ai-security), read 2026-11-16. - **harmonic-security-21** (vendor): Harmonic says it works without a network proxy or certificate changes, using a browser extension and Endpoint Agent plus OpenTelemetry and the local MCP Gateway.Quote: "without routing employee traffic through a network proxy or changing certificates"Source: [harmonic.security](https://www.harmonic.security/products/command), read 2026-11-16. - **harmonic-security-22** (observed): Harmonic's Protect Agent documentation says the agent reports desktop AI usage including connections to known AI network hosts.Quote: "Real-time reporting of desktop AI usage , including connections to known AI network hosts"Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/protect-agent/installation-and-deployment/frequently-asked-questions), read 2026-11-16. - **harmonic-security-23** (none-found): None found on 2026-11-16: email or billing records as a discovery source. Looked on the pricing page, Explore page and docs index.Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-24** (vendor): Harmonic says most teams see their first insight in about 30 minutes.Quote: "Most teams see first insight in about 30 minutes."Source: [harmonic.security](https://www.harmonic.security/solutions/shadow-ai-detection), read 2026-11-16. - **harmonic-security-25** (vendor): Harmonic says deployment takes minutes through Intune, JAMF or Kandji.Quote: "Deployment in minutes via Intune, JAMF, or Kandji."Source: [harmonic.security](https://www.harmonic.security/solutions/shadow-ai-detection), read 2026-11-16. - **harmonic-security-26** (observed): Harmonic's docs list supported operating systems for the Protect Agent.Quote: "Windows 11 macOS 15 (Sequoia), 26 (Tahoe) and 27 (Golden Gate) Ubuntu 22LTS, 24LTS and 26LTS Rocky Linux 9 and 10"Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/protect-agent/installation-and-deployment/frequently-asked-questions), read 2026-11-16. - **harmonic-security-27** (vendor): Harmonic's Explore page says that after deploying the browser extension a complete AI tool map appears within days.Quote: "within days you’ll have a complete map of the AI tools your workforce actually uses"Source: [harmonic.security](https://www.harmonic.security/products/explore), read 2026-11-16. - **harmonic-security-28** (vendor): Harmonic's pricing page says Guide warns, nudges or blocks inline in under 200 milliseconds.Quote: "Inline coaching that warns, nudges, or blocks in under 200 milliseconds"Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-29** (vendor): Harmonic says it can restrict personal or free accounts for corporate work.Quote: "Granular controls let you restrict personal or free accounts for corporate work, without cutting off the tools people genuinely need."Source: [harmonic.security](https://www.harmonic.security/solutions/shadow-ai-detection), read 2026-11-16. - **harmonic-security-30** (observed): Harmonic's webhook docs list alert and intervention events that can feed external workflows such as ticketing or training assignment.Quote: "Triggered when a new detection alert is raised."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/portal-guides/integration-guides/configure-a-webhook), read 2026-11-16. - **harmonic-security-31** (vendor): Harmonic's pricing page lists app tiers of approved, monitored and unsanctioned (Guide adds warning banners).Quote: "Organize tools into approved, monitored, and unsanctioned categories"Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-32** (observed): Harmonic's MCP access-control docs say a user's attempt to use a server they lack permission for is blocked and creates an access request for an administrator.Quote: "The attempt is blocked by the Harmonic MCP Gateway"Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/portal-guides/managing-your-organization/configure-mcp-server-access-controls), read 2026-11-16. - **harmonic-security-33** (observed): Harmonic's docs say every application gets a risk profile across four standalone risk factors.The exact quote is on the source page.Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/portal-guides/applications/application-risk-scoring-framework), read 2026-11-16. - **harmonic-security-34** (vendor): Harmonic says each tool has a profile of who uses it, how often, and whether the vendor trains on user inputs.Quote: "Each tool comes with a detailed profile covering who's using it, how often, and whether the vendor trains on user inputs."Source: [harmonic.security](https://www.harmonic.security/products/explore), read 2026-11-16. - **harmonic-security-35** (vendor): Harmonic's pricing page lists a full prompt audit log that can be streamed to a data lake (Command and Guide tiers show it as an add-on).Quote: "Full prompt audit log Stream all interaction data to your data lake for custom analysis"Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-36** (observed): Harmonic's docs say data from Applications, Employees, Alerts and Usage Explorer can be exported as CSV.Quote: "Export data from Applications, Employees, Alerts, and Usage Explorer as CSV files."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/llms.txt), read 2026-11-16. - **harmonic-security-37** (observed): Harmonic's product documentation is public without sign-in at docs.harmonicsecurity.app.Quote: "Browse by topic or search across all Harmonic docs."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/), read 2026-11-16. - **harmonic-security-38** (none-found): None found on 2026-11-16: a published price. The pricing page lists Explore, Guide and Command tiers with features but no amounts.Source: [harmonic.security](https://www.harmonic.security/pricing), read 2026-11-16. - **harmonic-security-39** (vendor): Harmonic publishes research: its AI Usage Index report (May 2026) analyzed 1.9 million classified AI-session minutes.Quote: "Harmonic Security analyzed 1.9 million classified AI-session minutes, tagging each conversation as personal, business, or ambiguous."Source: [harmonic.security](https://www.harmonic.security/resources/ai-usage-index-report-2026), read 2026-11-16. - **harmonic-security-51** (observed): Harmonic's docs say the local MCP Gateway is deployed by installing the Protect Agent, not the gateway directly.Quote: "Deployment of the MCP Gateway is done by installing the Harmonic Protect Agent rather than direct deployment of the gateway itself."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/mcp-gateway/local-mcp-gateway/installation-and-deployment), read 2026-11-16. - **harmonic-security-59** (observed): Harmonic's docs say coding-agent inline controls cover prompts and tool calls for Claude Code, Codex, Cursor, VS Code, GitHub Copilot CLI and others.Quote: "Coding agent inline controls covers user prompts and tool calls."Source: [docs.harmonicsecurity.app](https://docs.harmonicsecurity.app/portal-guides/detection-and-response/sensitive-data-controls-for-coding-agents-and-desktop-applications), read 2026-11-16. ## News mentioning Harmonic Security - [Harmonic Security raises $50 million Series B](/news/harmonic-security-raises-50-million-series-b/) ## Related pages - [All vendor profiles](/vendors/) - [Grip Security profile](/vendors/grip-security/) - [LayerX (Akamai Workforce Protector) profile](/vendors/layerx/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # LayerX (Akamai Workforce Protector) for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/layerx/ Updated: 2026-11-16 LayerX is now sold as Akamai Workforce Protector. A browser extension, with a management console and cloud intelligence, discovers shadow AI apps, AI browsers, extensions and desktop tools, and applies warn, redact, block and guide actions, including blocking risky extensions. Akamai says it works without proxies or network changes, and four browsers are named. Reported coverage includes tools and identities in use, data in prompts and records for regulators. Marketing pages are readable. None found on 16 Nov 2026 for discovery of MCP servers, embedded SaaS AI or OAuth-granted AI apps, for a published price, public documentation, research or named customers. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Browser-native discovery of shadow AI apps, AI browsers, extensions and desktop tools with warn, redact and block actions, deployed without proxies (facts 06, 07, 20, 22). **Weakness:** None found on 2026-11-16 for MCP servers, embedded SaaS AI or OAuth-granted AI apps, a published price, public docs, or named customers (facts 11, 23, 28, 29, 31). ## Scores and evidence LayerX (Akamai Workforce Protector) is 15th in the overall ranking at 45.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **LayerX (Akamai Workforce Protector): score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 5 of 10 | 25 | Web apps, desktop tools, agents, browser extensions and named agentic browsers. None found for MCP servers, embedded SaaS AI or OAuth apps. | [layerx-06](#fact-layerx-06), [layerx-07](#fact-layerx-07), [layerx-08](#fact-layerx-08), [layerx-09](#fact-layerx-09), [layerx-11](#fact-layerx-11) | | Where it looks | 4 of 10 | 20 | Browser extension and desktop visibility, with unnamed IdP integrations. None found for network, email or billing sources. | [layerx-12](#fact-layerx-12), [layerx-13](#fact-layerx-13), [layerx-14](#fact-layerx-14), [layerx-15](#fact-layerx-15) | | Time to a first inventory | 5 of 10 | 15 | Extension on existing browsers with four browsers named and no proxy. None found for a time to first inventory or OS list. | [layerx-16](#fact-layerx-16), [layerx-17](#fact-layerx-17), [layerx-18](#fact-layerx-18), [layerx-19](#fact-layerx-19) | | What it can do about a finding | 6 of 10 | 15 | Warn, redact, block and guide, and blocking risky extensions. None found for OAuth revocation or approval workflow. | [layerx-20](#fact-layerx-20), [layerx-21](#fact-layerx-21), [layerx-22](#fact-layerx-22), [layerx-23](#fact-layerx-23) | | Risk context and reporting | 5 of 10 | 10 | Tools and identities in use, data in prompts, risk analysis and records for regulators. None found for a training rating or export format. | [layerx-24](#fact-layerx-24), [layerx-25](#fact-layerx-25), [layerx-26](#fact-layerx-26), [layerx-08](#fact-layerx-08) | | Transparency | 1 of 10 | 10 | Marketing pages readable. None found for documentation, price or research. | [layerx-27](#fact-layerx-27), [layerx-28](#fact-layerx-28), [layerx-29](#fact-layerx-29) | | Proof | 5 of 10 | 5 | Acquisition and funding reported by SecurityWeek and a Gartner sample vendor listing. None found for named customers. | [layerx-03](#fact-layerx-03), [layerx-04](#fact-layerx-04), [layerx-30](#fact-layerx-30), [layerx-31](#fact-layerx-31) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 15th | 45.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 15th | 47.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 14th | 46.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 15th | 47.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 15th | 46.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 14th | 50.0 | ## Product, deployment, price and owner - Product assessed: Akamai Workforce Protector (formerly LayerX): browser extension, management console and cloud intelligence - Deployment: Browser extension plus management console and cloud intelligence; no proxy or network changes (Akamai claim) - Time to deploy: none found on 2026-11-16 - Price: none found on 2026-11-16 - Owner: Akamai Technologies, Inc. - Website: layerxsecurity.com ## Facts and sources - **layerx-01** (vendor): Akamai says LayerX is now Akamai Workforce Protector, part of Akamai's security portfolio.Quote: "LayerX is now Akamai Workforce Protector."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-02** (vendor): Akamai announced it completed its acquisition of LayerX; the release is dated 2026-07-02.Quote: "announces the company has completed its acquisition of secure enterprise browser provider and AI usage control leader LayerX."Source: [akamai.com](https://www.akamai.com/newsroom/press-release/akamai-completes-acquisition-of-secure-enterprise-browser-provider-layerx), read 2026-11-16. - **layerx-03** (third-party): SecurityWeek reported on 2026-05-14 that Akamai agreed to buy LayerX for about $205 million in cash.Quote: "agreed to acquire LayerX, a company specializing in AI and browser security, for roughly $205 million in an all-cash deal."Source: [securityweek.com](https://www.securityweek.com/akamai-to-acquire-ai-and-browser-security-firm-layerx-for-205-million/), read 2026-11-16. - **layerx-04** (third-party): SecurityWeek reported LayerX had raised $45 million in total before the deal.Quote: "a total of $45 million, has developed a security platform that provides real-time visibility and control over user and agentic activities across browsers, applications, and IDEs."Source: [securityweek.com](https://www.securityweek.com/akamai-to-acquire-ai-and-browser-security-firm-layerx-for-205-million/), read 2026-11-16. - **layerx-05** (vendor): Akamai's announcement of intent to acquire names Or Eshed as LayerX CEO and co-founder, joining Akamai's Zero Trust organization with co-founder David Vaisbrud.Quote: "said Or Eshed, Chief Executive Officer and Co-Founder, LayerX."Source: [akamai.com](https://www.akamai.com/newsroom/press-release/akamai-technologies-announces-intent-to-acquire-layerx-advancing-its-workforce-security-strategy-with-ai-usage-control), read 2026-11-16. - **layerx-06** (vendor): Akamai says Workforce Protector discovers shadow AI across web apps, browsers and desktop tools.Quote: "Discover shadow AI across web applications, browsers, and desktop tools"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-07** (vendor): Akamai says it discovers AI applications, browsers, desktop tools and agents.Quote: "Discover AI applications, browsers, desktop tools, and agents so users can adopt AI safely and responsibly."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-08** (vendor): Akamai says it monitors browser extensions and scores their risk, behavior and permissions.Quote: "Monitor browser extensions and reduce extension-based risk"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-09** (vendor): Akamai says it covers agentic browsers by name (ChatGPT Atlas, Perplexity Comet, Dia).Quote: "including agentic browsers such as ChatGPT Atlas, Perplexity Comet, and Dia."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-10** (vendor): Akamai says it can tell when a user reaches an AI tool with a personal rather than corporate identity.Quote: "AI usage control can identify when someone authenticates to an AI tool with a personal rather than a corporate identity"Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-11** (none-found): None found on 2026-11-16: MCP servers, AI features embedded in approved SaaS, or AI apps granted OAuth access as discovery targets, on the Akamai Workforce Protector and AI usage control pages (the former layerxsecurity.com pages redirect there).Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-12** (vendor): Akamai says the product combines a browser extension, desktop visibility, a management console and cloud intelligence.Quote: "Workforce Protector combines browser-native protection, desktop visibility, centralized policy management, and cloud intelligence to secure how users and AI agents work."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-13** (vendor): Akamai says it observes AI use in desktop applications and agent workflows outside the browser as well.Quote: "including desktop applications and agent-driven workflows that never touch a browser session."Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-14** (vendor): Akamai says it works with existing identity and SIEM tools and lists IdP, SIEM, file-labeling, ticketing and MDM integrations without naming products on the page.Quote: "Existing identity, SIEM, and DLP investments stay in place and gain visibility they did not previously have."Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-15** (vendor): Akamai says it partners with Google Workspace and Chrome to discover shadow AI usage and manage extension risk.Quote: "Partners with Google Workspace and Chrome to manage browser extension risk, discover shadow AI usage, and prevent sensitive data exfiltration."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-16** (vendor): Akamai says deployment needs no proxies, network redesign or browser replacement.Quote: "Deploy quickly without proxies, network redesign, or browser replacement"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-17** (vendor): Akamai says it covers Chrome, Edge, Firefox and Safari and works on managed and unmanaged devices.Quote: "Apply controls across Chrome, Edge, Firefox, and Safari"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-18** (vendor): Akamai says it installs as an extension on browsers already in use; managed and unmanaged devices are covered.Quote: "Secure managed and unmanaged devices without changing user workflows"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-19** (none-found): None found on 2026-11-16: a published time to first AI inventory or a list of supported operating systems on the Akamai Workforce Protector page.Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-20** (vendor): Akamai says policies can warn, redact, block or guide users based on context and risk.Quote: "Apply adaptive controls that warn, redact, block, or guide user actions based on context and risk."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-21** (vendor): Akamai says enforcement happens before data leaves the user.Quote: "Enforcement happens before data leaves, not after a log is reviewed."Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-22** (vendor): Akamai says it blocks malicious or overprivileged browser extensions using a risk database.Quote: "to instantly block malicious or overprivileged third-party browser plug-ins."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-23** (none-found): None found on 2026-11-16: OAuth grant revocation or an approval workflow for AI tools on the Akamai Workforce Protector and AI usage control pages.Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-24** (vendor): Akamai says it shows which AI tools and identities are in use and what data moves into prompts and uploads.Quote: "It shows you which AI tools and identities are in use, what data moves into prompts and uploads"Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-25** (vendor): Akamai says cloud intelligence supplies risk analysis for applications, identities, extensions and AI activity.Quote: "Delivers risk analysis and intelligence for applications, identities, browser extensions, and AI activity."Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-26** (vendor): Akamai says the product records which AI tools were used, by which identities, with what data, for regulators.Quote: "AI usage control produces a record of which AI tools were used, under which identities, and what data was involved"Source: [akamai.com](https://www.akamai.com/solutions/ai-usage-control), read 2026-11-16. - **layerx-27** (vendor): Akamai's product and solution pages are readable without sign-in and offer a demo request.Quote: "Get a free demo to see how Workforce Protector helps security teams:"Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-28** (none-found): None found on 2026-11-16: a published price for Workforce Protector or LayerX; the former layerxsecurity.com/pricing returned 404 and the Akamai page only offers a demo.Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-29** (none-found): None found on 2026-11-16: public product documentation; docs.layerxsecurity.com redirected to a login page, and no published research was reachable on the pages read.Source: [docs.layerxsecurity.com](https://docs.layerxsecurity.com), read 2026-11-16. - **layerx-30** (third-party): A GlobeNewswire listing shows LayerX announced on 2025-08-14 that it was named a Sample Vendor in Gartner coverage of the Secure Enterprise Browser category (text truncated on the listing).Quote: "today announced its recognition as a Sample Vendor in both the Secure Enterprise..."Source: [globenewswire.com](https://www.globenewswire.com/search/keyword/LayerX), read 2026-11-16. - **layerx-31** (none-found): None found on 2026-11-16: named customers on the Akamai Workforce Protector pages or the acquisition coverage read.Source: [akamai.com](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. - **layerx-32** (none-found): None found on 2026-11-16: news dated 2026-08-18 or later about this product beyond the 2026-07-02 completion release; the GlobeNewswire LayerX listing shows its newest items from May 2026 or earlier.Source: [globenewswire.com](https://www.globenewswire.com/search/keyword/LayerX), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Harmonic Security profile](/vendors/harmonic-security/) - [Microsoft profile](/vendors/microsoft/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Microsoft for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/microsoft/ Updated: 2026-11-16 Microsoft covers shadow AI through three products that most Microsoft 365 customers already license in part. Defender for Cloud Apps compares network traffic against a catalog of more than 31,000 apps scored on 90-plus risk factors and lets admins unsanction an app. Defender for Endpoint discovers local AI agents, including Claude Code, Cursor, Codex CLI and GitHub Copilot, on Windows and macOS. Purview adds browser-side prompt insight and Endpoint DLP paste control. Microsoft's own tutorial says apps outside the catalog are not discovered by default, which matters for new AI tools. None found on 16 Nov 2026 for a published time to a first inventory. Agent 365 is listed at $15.00 per user per month. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Defender for Cloud Apps scores a catalog of over 31,000 apps on 90+ risk factors from endpoint, firewall and proxy feeds, with unsanction and block actions (microsoft-02, microsoft-03, microsoft-04). **Weakness:** Apps outside the catalog are not discovered by default, and none found on 2026-11-16 for a published time to a first AI inventory (microsoft-05, microsoft-60). ## Scores and evidence Microsoft is 2nd in the overall ranking at 66.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Microsoft: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 8 of 10 | 25 | Catalog AI apps, OAuth apps, named local coding agents, IDE extensions, desktop assistants and MCP configurations, with 60 agent types confirmed by third parties. None found for embedded SaaS AI; uncatalogued apps are missed. | [microsoft-14](#fact-microsoft-14), [microsoft-11](#fact-microsoft-11), [microsoft-06](#fact-microsoft-06), [microsoft-07](#fact-microsoft-07), [microsoft-54](#fact-microsoft-54), [microsoft-65](#fact-microsoft-65), [microsoft-05](#fact-microsoft-05) | | Where it looks | 7 of 10 | 20 | Endpoint, firewall and proxy logs with a named proxy integration, browser extension and OAuth app governance; vendor-only. | [microsoft-02](#fact-microsoft-02), [microsoft-08](#fact-microsoft-08), [microsoft-06](#fact-microsoft-06), [microsoft-09](#fact-microsoft-09), [microsoft-40](#fact-microsoft-40) | | Time to a first inventory | 6 of 10 | 15 | Discovery starts automatically after device onboarding, OS list published, DSPM needs 24 hours. None found for a time to first inventory. | [microsoft-12](#fact-microsoft-12), [microsoft-13](#fact-microsoft-13), [microsoft-61](#fact-microsoft-61), [microsoft-60](#fact-microsoft-60) | | What it can do about a finding | 6 of 10 | 15 | Unsanction and block via proxy, restrict OAuth apps, DLP paste control and redirection. None found for coaching or tickets. | [microsoft-04](#fact-microsoft-04), [microsoft-06](#fact-microsoft-06), [microsoft-10](#fact-microsoft-10), [microsoft-67](#fact-microsoft-67) | | Risk context and reporting | 6 of 10 | 10 | 31,000-app catalog with 90+ risk factors, usage and user drill-down and agent risk levels. None found for an export. | [microsoft-03](#fact-microsoft-03), [microsoft-06](#fact-microsoft-06), [microsoft-18](#fact-microsoft-18), [microsoft-01](#fact-microsoft-01) | | Transparency | 6 of 10 | 10 | Public documentation and a published price for one component. None found for published research. | [microsoft-52](#fact-microsoft-52), [microsoft-49](#fact-microsoft-49), [microsoft-51](#fact-microsoft-51) | | Proof | 5 of 10 | 5 | SiliconANGLE covers the launch by a large company. None found for named customers or analyst coverage for this product. | [microsoft-64](#fact-microsoft-64), [microsoft-58](#fact-microsoft-58) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 2nd | 66.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 3rd | 66.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 2nd | 67.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 2nd | 69.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | joint 2nd | 67.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 2nd | 65.0 | ## Product, deployment, price and owner - Product assessed: Microsoft Defender for Cloud Apps (cloud discovery, app governance), Defender for Endpoint local AI agent discovery, Microsoft Purview DSPM for AI and Endpoint DLP - Deployment: Cloud-managed Defender for Endpoint client onboarding (Intune or PowerShell for runtime protection); API-based Content Safety; Purview browser extension; Entra and Defender portal controls - Time to deploy: none found on 2026-11-16 (discovery starts automatically after device onboarding; Purview policies need at least 24 hours) - Price: Agent 365 is listed at $15.00 per user per month, paid yearly; Content Safety Standard shows a dash; Defender for Endpoint and Cloud Apps prices none found on 2026-11-16 on readable pages. - Owner: Microsoft Corporation - Website: learn.microsoft.com ## Facts and sources - **microsoft-01** (vendor): Microsoft says Defender for Cloud Apps discovers cloud and SaaS apps from network traffic against an app catalog and ranks each app's risk.Quote: "uses data based on an assessment of network traffic and an extensive app catalog to identify apps accessed by users across your organization"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/what-is-defender-for-cloud-apps), read 2026-11-16. - **microsoft-02** (vendor): Microsoft documents three cloud discovery feeds: the Defender for Endpoint integration on Windows 10 and 11 devices, a log collector on firewalls and proxies, and proxy integrations such as Zscaler.Quote: "This native integration enables you to immediately start collecting data on cloud traffic across your Windows 10 and Windows 11 devices, on and off your network."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. - **microsoft-03** (vendor): Microsoft says its app catalog holds over 31,000 apps scored on more than 90 risk factors, with per-app usage and user drill-down.Quote: "The Defender for cloud app catalog includes over 31,000 apps that are assessed using over 90 risk factors."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. - **microsoft-04** (vendor): Microsoft says admins can unsanction a discovered app and have it blocked natively through their proxy or firewall appliance.Quote: "to unsanction these apps in order to block them natively using your proxy or firewall appliance"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. - **microsoft-05** (vendor): The same tutorial says apps outside the catalog cannot be discovered by default, a limit for new AI tools.Quote: "By default, Defender for Cloud Apps can't discover apps that aren't in the catalog."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. - **microsoft-06** (vendor): Microsoft says app governance covers OAuth-enabled apps registered in Entra ID, Google and Salesforce, showing the permissions they hold and who granted access, and supports blocking or restricting risky apps.Quote: "see which user-installed OAuth applications have access to data on Microsoft 365, Google Workspace, and Salesforce"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance), read 2026-11-16. - **microsoft-07** (vendor): Microsoft Purview says AI apps seen through browser activity and filed as Generative AI in the Defender for Cloud Apps catalog include third-party tools such as ChatGPT, Google Gemini and DeepSeek.Quote: "Other AI apps that are detected through browser activity and categorized as "Generative AI" in the Defender for Cloud Apps catalog."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/ai-microsoft-purview), read 2026-11-16. - **microsoft-08** (vendor): Microsoft says third-party AI site insights in DSPM for AI need the Purview browser extension and onboarded devices as prerequisites.Quote: "Install Microsoft Purview browser extension A prerequisite for third-party AI sites."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/dspm-for-ai), read 2026-11-16. - **microsoft-09** (vendor): Microsoft documents a DSPM for AI recommendation that detects sensitive information shared with AI apps in browsers, applications, APIs and add-ins through a Secure Access Service Edge or Security Service Edge integration.Quote: "uses network data security to detect sensitive info types shared with AI apps in browsers, applications, APIs, add-ins, and more"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/dspm-for-ai), read 2026-11-16. - **microsoft-10** (vendor): Microsoft documents Endpoint DLP detection of pasting content to restricted service domains in supported browsers, evaluated on the pasted content.Quote: "Detects when a user attempts to paste content to a restricted service domain. Evaluation is performed on the content that is being pasted."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about), read 2026-11-16. - **microsoft-11** (vendor): Microsoft says Defender for Endpoint discovers supported local AI agents on Windows devices, including coding agents, IDE extensions, desktop assistants and agent platforms, and lists them in an AI agent inventory.The exact quote is on the source page.Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint), read 2026-11-16. - **microsoft-12** (vendor): Microsoft says local agent discovery needs no extra deployment beyond onboarding to Defender for Endpoint and starts automatically; Plan 2 is the minimum license.Quote: "You don't need additional deployment, configuration, or scripts beyond the device onboarding requirements. If the device meets all of the prerequisites, agent discovery begins automatically."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents), read 2026-11-16. - **microsoft-13** (vendor): Microsoft lists Windows, macOS, Linux, Android and iOS as Defender for Endpoint operating systems; local agent discovery is documented for Windows and macOS.Quote: "Microsoft Defender for Endpoint supports the following operating systems: Windows, macOS, Linux, Android, and iOS."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/microsoft-defender-endpoint), read 2026-11-16. - **microsoft-14** (vendor): Microsoft's support matrix names the local agents discovered, such as Claude Code, Claude Desktop, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf and Kiro, and marks which also report MCP server configuration.Quote: "Discovers MCP column indicates that Defender reports remote MCP server data in McpServers and local MCP server data in RawAgentInfo.localAgentMetadata.localMcps"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix), read 2026-11-16. - **microsoft-15** (vendor): On 2026-10 release notes, Microsoft says local AI agent discovery is generally available on Windows and macOS and reports each agent's configured MCP servers.Quote: "Discover supported local AI agents and their configured MCP servers on onboarded Windows and macOS devices"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint), read 2026-11-16. - **microsoft-16** (vendor): Microsoft says the AI agent inventory lists, per local agent, the remote and local MCP servers with name, type and endpoint.Quote: "MCP servers lists the MCP servers configured for the agent, with the name, type, and endpoint of each server."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents), read 2026-11-16. - **microsoft-18** (vendor): Microsoft says the agent risk level, indicators and recommendations for discovered local agents need Microsoft 365 E7 or Agent 365 together with Defender for Endpoint Plan 2.Quote: "View the risk level, risk indicators, and security recommendations for discovered local AI agents"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents), read 2026-11-16. - **microsoft-25** (vendor): Microsoft says runtime protection is configured in audit or block mode through local PowerShell or organization-wide Intune.Quote: "For local PowerShell and organization-wide Microsoft Intune configuration options"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-runtime-protection-overview), read 2026-11-16. - **microsoft-38** (vendor): Microsoft says Agent 365 is generally available for the Commercial segment per user, works best with Microsoft E5, and needs at least one licensed user.Quote: "Microsoft Agent 365 is generally available for the Commercial segment on a per user basis."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/microsoft-agent-365/overview), read 2026-11-16. - **microsoft-40** (none-found): None found on 2026-11-16: no Microsoft page read describes a dedicated AI browser or AI extension agent control in Defender; the browser-side evidence is the Purview browser extension and Endpoint DLP paste control.Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/dspm-for-ai), read 2026-11-16. - **microsoft-49** (vendor): Microsoft lists Agent 365 at $15.00 per user per month, paid yearly, on its product page.Quote: "$15.00 user/month, paid yearly (Annual commitment)"Source: [microsoft.com](https://www.microsoft.com/en-us/microsoft-agent-365), read 2026-11-16. - **microsoft-50** (vendor): Microsoft's Azure AI Content Safety pricing page shows a free tier of 5,000 text records per month and a Standard tier whose price displays as a dash.The exact quote is on the source page.Source: [azure.microsoft.com](https://azure.microsoft.com/en-us/pricing/details/content-safety/), read 2026-11-16. - **microsoft-51** (none-found): None found on 2026-11-16: no price for Defender for Cloud Apps or Defender for Endpoint standalone is readable on the learn pages; they point to Microsoft 365 E5 inclusion and separate pricing pages that returned 403 to a script.Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/microsoft-defender-endpoint), read 2026-11-16. - **microsoft-52** (vendor): Microsoft documentation on learn.microsoft.com is readable without sign-in, with last-updated dates (Agent 365 overview 2026-11-12).Quote: "Last updated on 2026-11-12"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/microsoft-agent-365/overview), read 2026-11-16. - **microsoft-54** (third-party): On 2026-10-20, Help Net Security reported Microsoft is expanding discovery to 35 more local agent types for a total of 60, including Claude Code, OpenAI Codex and GitHub Copilot.Quote: "Microsoft Security is expanding discovery to 35 additional types of local agents, bringing the total supported to 60"Source: [helpnetsecurity.com](https://www.helpnetsecurity.com/2026/10/20/microsofts-ai-agents-threats-guardian-agents/), read 2026-11-16. - **microsoft-56** (third-party): On 2026-10-20, Help Net Security quoted Satya Nadella, CEO of Microsoft, on treating frontier models as insider risks, and said Project Perception shares signals with Agent 365.Quote: "Satya Nadella , CEO of Microsoft, wrote"Source: [helpnetsecurity.com](https://www.helpnetsecurity.com/2026/10/20/microsofts-ai-agents-threats-guardian-agents/), read 2026-11-16. - **microsoft-58** (none-found): None found on 2026-11-16: no named customer list or analyst page for shadow AI discovery, MCP security or AI runtime security was read on the Microsoft pages fetched.Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint), read 2026-11-16. - **microsoft-59** (vendor): Microsoft owns and sells these products itself: Defender for Endpoint is described as the endpoint security pillar of Microsoft Defender.Quote: "As the endpoint security pillar of Microsoft Defender , Defender for Endpoint feeds endpoint signals into the unified Defender portal."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-endpoint/microsoft-defender-endpoint), read 2026-11-16. - **microsoft-60** (none-found): None found on 2026-11-16: no published time to a first AI inventory for Defender for Cloud Apps cloud discovery was found on the shadow IT tutorial; only that Defender for Endpoint integration starts collecting immediately.Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. - **microsoft-61** (vendor): Microsoft Purview says DSPM for AI policies need at least 24 hours to collect data before results appear.Quote: "Allow at least 24 hours for these new policies to collect data to display the results in Data Security Posture Management for AI"Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/dspm-for-ai), read 2026-11-16. - **microsoft-64** (third-party): On 2026-10-19, SiliconANGLE covered the launch and named Satya Nadella as Microsoft Chairman and Chief Executive and Hayete Gallot as executive vice president of Microsoft Security.Quote: "Microsoft Chairman and Chief Executive Satya Nadella laid out the thinking behind the controls in an Oct. 10 essay"Source: [siliconangle.com](https://siliconangle.com/2026/10/19/microsoft-targets-the-agentic-insider-with-new-controls-for-rogue-ai-agents/), read 2026-11-16. - **microsoft-65** (third-party): On 2026-10-19, SiliconANGLE reported Microsoft's discovery tools can now spot 60 types of local agents on employee machines after 35 were added.Quote: "discovery tools can now spot 60 types of local agents on employee machines after 35 were added."Source: [siliconangle.com](https://siliconangle.com/2026/10/19/microsoft-targets-the-agentic-insider-with-new-controls-for-rogue-ai-agents/), read 2026-11-16. - **microsoft-67** (vendor): Microsoft says Endpoint DLP blocks an upload from an unallowed browser and redirects the user to Microsoft Edge, a browser-side control that is not specific to AI agents.Quote: "If they're using an unallowed browser, the upload activity is blocked and the user is redirected to use Microsoft Edge."Source: [learn.microsoft.com](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [LayerX (Akamai Workforce Protector) profile](/vendors/layerx/) - [Netskope profile](/vendors/netskope/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Netskope for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/netskope/ Updated: 2026-11-16 Netskope reads shadow AI from its secure web gateway traffic and from the Netskope Client (version 138 or later) on endpoints. Its documentation, readable without sign-in, covers browser, editor and desktop AI extensions, and the company says the AI Command Center also finds local models, MCP servers and AI embedded in SaaS, with per-app risk scores that include training on customer data. Actions include coaching, redaction, blocking and MCP server access policy. Discovery depends on an existing Netskope gateway or client deployment. None found on 16 Nov 2026 for discovery of OAuth-granted AI apps, a published price, a time to a first inventory or an inventory export. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Discovers AI web apps, embedded SaaS AI, MCP servers, local models and AI extensions from network and endpoint sources, with per-app risk ratings and coaching or blocking (netskope-04, netskope-05, netskope-08, netskope-17, netskope-23). **Weakness:** None found on 2026-11-16 for OAuth-granted AI app discovery, a published price, a time to first inventory or an inventory export; discovery depends on an existing Netskope gateway or client deployment (netskope-07, netskope-16, netskope-22, netskope-26, netskope-28). ## Scores and evidence Netskope is 3rd in the overall ranking at 65.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Netskope: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 8 of 10 | 25 | Docs observe discovery of browser, editor and desktop AI extensions; claims add local agents, LLMs, MCP servers, AI apps and embedded SaaS AI, with MCP discovery confirmed by SiliconANGLE. None found for OAuth-granted AI apps. | [netskope-04](#fact-netskope-04), [netskope-05](#fact-netskope-05), [netskope-06](#fact-netskope-06), [netskope-08](#fact-netskope-08), [netskope-09](#fact-netskope-09), [netskope-34](#fact-netskope-34), [netskope-07](#fact-netskope-07) | | Where it looks | 7 of 10 | 20 | Docs observe the endpoint client, the secure web gateway and identity-provider identities; API-based protection is claimed. None found for email or billing records. | [netskope-10](#fact-netskope-10), [netskope-11](#fact-netskope-11), [netskope-12](#fact-netskope-12), [netskope-09](#fact-netskope-09) | | Time to a first inventory | 4 of 10 | 15 | No new infrastructure if the platform is in place, but docs require the gateway and a recent client. None found for a time to first inventory or an OS and browser list. | [netskope-14](#fact-netskope-14), [netskope-15](#fact-netskope-15), [netskope-10](#fact-netskope-10), [netskope-16](#fact-netskope-16) | | What it can do about a finding | 6 of 10 | 15 | Coaching, redaction, blocking, MCP server access policy and in-app action controls, vendor documented. None found for OAuth revocation or ticket routing. | [netskope-17](#fact-netskope-17), [netskope-18](#fact-netskope-18), [netskope-19](#fact-netskope-19), [netskope-20](#fact-netskope-20), [netskope-21](#fact-netskope-21), [netskope-22](#fact-netskope-22) | | Risk context and reporting | 6 of 10 | 10 | Risk ratings include training on customer data and an inventory shows risk scores and usage. None found for an exportable inventory or report. | [netskope-23](#fact-netskope-23), [netskope-24](#fact-netskope-24), [netskope-25](#fact-netskope-25), [netskope-26](#fact-netskope-26) | | Transparency | 6 of 10 | 10 | Documentation observed readable without sign-in and a published research report. None found for a price. | [netskope-27](#fact-netskope-27), [netskope-29](#fact-netskope-29), [netskope-28](#fact-netskope-28) | | Proof | 9 of 10 | 5 | Public company with IPO and funding reported by SecurityWeek, a named customer and a Gartner Leader placement. | [netskope-30](#fact-netskope-30), [netskope-31](#fact-netskope-31), [netskope-01](#fact-netskope-01), [netskope-32](#fact-netskope-32), [netskope-33](#fact-netskope-33) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 3rd | 65.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 2nd | 67.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 4th | 63.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 2nd | 69.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | joint 2nd | 67.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 2nd | 65.0 | ## Product, deployment, price and owner - Product assessed: Netskope Skylight AI Command Center, with Netskope Skylight GenAI App Security, Agentic Broker and the Netskope Client endpoint AI discovery, on the Netskope One platform - Deployment: Runs on the Netskope One platform using Next-Gen SWG traffic and the Netskope Client (version 138 or later) for endpoint AI discovery - Time to deploy: none found on 2026-11-16 - Price: none found on 2026-11-16 - Owner: Netskope, Inc. (Nasdaq: NTSK), standalone public company - Website: netskope.com ## Facts and sources - **netskope-01** (vendor): Netskope is a publicly traded company (Nasdaq: NTSK) and describes itself as a leader in security and networking for the cloud and AI era.Quote: "Netskope (NASDAQ: NTSK), a leader in modern security and networking for the cloud and AI era"Source: [netskope.com](https://www.netskope.com/press-releases/netskope-enables-security-teams-to-stop-risky-ai-agent-actions-before-they-execute), read 2026-11-16. - **netskope-02** (third-party): Sanjay Beri is co-founder and Chief Executive of Netskope, per SiliconANGLE.Quote: "co-founder and Chief Executive Sanjay Beri"Source: [siliconangle.com](https://siliconangle.com/2026/03/11/netskope-launches-netskope-one-ai-security-suite-protect-agentic-ai-enterprise-models), read 2026-11-16. - **netskope-03** (vendor): On 2026-09-15 Netskope said its AI security portfolio, formerly Netskope One AI Security, is now named Netskope Skylight, which includes AI Command Center, AI Guardrails, AI Gateway, Agentic Broker and AI Red Teaming.Quote: "Formerly Netskope One AI Security, Netskope Skylight is the new family name for the suite of Netskope AI security products"Source: [netskope.com](https://www.netskope.com/press-releases/netskope-enables-security-teams-to-stop-risky-ai-agent-actions-before-they-execute), read 2026-11-16. - **netskope-04** (vendor): Netskope says endpoint visibility identifies AI running locally, including known AI agents, LLMs and MCP servers.Quote: "endpoint visibility identifies AI running locally including known AI agents, LLMs and MCP servers"Source: [netskope.com](https://www.netskope.com/products/ai-products), read 2026-11-16. - **netskope-05** (observed): Netskope documents that AI Command Center discovers browser, editor and desktop AI extensions as agents, sourced from the Netskope Client endpoint AI discovery.Quote: "AI-enabled extensions that run inside a browser, code editor, or desktop, such as browser extensions, editor extensions, and desktop extensions."Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-06** (vendor): Netskope says AI Command Center continuously discovers AI apps, AI embedded in SaaS apps, and MCP servers.Quote: "continuous discovery of AI apps, AI embedded within SaaS apps, and the MCP servers enabling autonomous agents"Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-07** (none-found): None found on 2026-11-16: looked for discovery of AI apps granted OAuth access to Google Workspace or Microsoft 365 on the AI Command Center page, its docs page and the generative AI page; no OAuth grant discovery is described.Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-08** (vendor): Netskope says AI Command Center identifies AI features embedded in managed and unmanaged SaaS apps, with risk scores.Quote: "identifies embedded genAI features (such as connections to MCP servers) within SaaS apps (both managed and unmanaged)"Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-09** (third-party): SiliconANGLE reported on 2025-12-01 that Netskope One can automatically discover MCP servers and MCP-enabled clients on enterprise networks.Quote: "automatically discover and identify both MCP servers and MCP-enabled clients operating on enterprise networks"Source: [siliconangle.com](https://siliconangle.com/2025/12/01/netskope-adds-mcp-security-controls-protect-enterprise-ai-agents), read 2026-11-16. - **netskope-10** (observed): Netskope docs say endpoint AI discovery of MCP servers, models and agents must be enabled and requires Netskope Client version 138 or later.Quote: "Endpoint AI discovery must be enabled. Requires Netskope Client version 138 or later."Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-11** (observed): Netskope docs say AI application and identity visibility comes from the Next-Gen Secure Web Gateway, with known identities being users authenticated through the identity provider.Quote: "Known identities are users authenticated through your Identity Provider (IdP)."Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-12** (vendor): Netskope says it detects and blocks threats in generative AI apps with inline and API-based protection.Quote: "detect and block threats in real time with inline and API-based protection"Source: [netskope.com](https://www.netskope.com/products/securing-generative-ai), read 2026-11-16. - **netskope-13** (vendor): Netskope says Skylight discovers remote MCP servers, clients, tools, resources and prompt requests in use in the organization.Quote: "Continuous discovery of remote the MCP servers, clients, tools, resources, and prompt requests in use within your organization"Source: [netskope.com](https://www.netskope.com/products/agentic-broker), read 2026-11-16. - **netskope-14** (vendor): Netskope says AI Command Center needs no new infrastructure and uses signals from the existing Netskope One platform.Quote: "It uses signals from across the Netskope One Platform to provide visibility into AI application and MCP server usage, risk, and policy violations."Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-15** (observed): Netskope docs say AI application discovery requires the Next-Gen Secure Web Gateway to be enabled and operational, so setup depends on an existing deployment.Quote: "Next-Gen Secure Web Gateway (NG-SWG) must be enabled and fully operational"Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-16** (none-found): None found on 2026-11-16: looked for a published time to first AI inventory and a list of supported operating systems and browsers on the AI Command Center docs page; neither is stated there.Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-17** (vendor): Netskope says real-time user coaching explains the risk and points to the approved option instead of an unexplained block.Quote: "Real-time user coaching explains the risk to the user at the most effective moment, and points to the approved option, instead of applying an unexplained block."Source: [netskope.com](https://www.netskope.com/products/ai-products), read 2026-11-16. - **netskope-18** (vendor): Netskope says redaction and blocking stop sensitive data before it reaches a model.Quote: "Redaction and blocking stop sensitive data before it reaches a model, agent actions are controlled, adversaries get blocked"Source: [netskope.com](https://www.netskope.com/products/ai-products), read 2026-11-16. - **netskope-19** (vendor): Netskope says policy can set access to cataloged public MCP servers, with a default block option for all public MCP server use.Quote: "Enforce policy controls over public MCP servers, determining access permissions for cataloged public MCP servers, with a default block option for all public MCP server use."Source: [netskope.com](https://www.netskope.com/products/agentic-broker), read 2026-11-16. - **netskope-20** (vendor): On 2026-09-15 Netskope announced Agent Action Control, with block, allow or alert policies by risk level and optional user coaching notifications.Quote: "Block, allow or alert based on low, medium, high or critical risk categories"Source: [netskope.com](https://www.netskope.com/press-releases/netskope-enables-security-teams-to-stop-risky-ai-agent-actions-before-they-execute), read 2026-11-16. - **netskope-21** (vendor): Netskope says teams can control specific actions inside AI applications such as uploading, downloading, copying and printing.Quote: "Teams can also control specific actions such as uploading, downloading, copying, printing, and more within AI applications to reduce risk."Source: [netskope.com](https://www.netskope.com/products/securing-generative-ai), read 2026-11-16. - **netskope-22** (none-found): None found on 2026-11-16: looked for revoking OAuth grants and routing a discovered AI tool to a ticket on the AI Command Center page; neither is described.Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-23** (vendor): Netskope says it rates risk for over 82,000 apps, including whether they use customer data for training or share data with third parties.Quote: "including if they use customer data for training, share data with third-parties, meet specific compliance requirements, and more"Source: [netskope.com](https://www.netskope.com/products/securing-generative-ai), read 2026-11-16. - **netskope-24** (vendor): Netskope says AI Command Center keeps an inventory of AI assets with risk scores, usage activity, threats and policy violations.Quote: "Maintain an inventory of AI assets with risk scores, usage activity, threats, and any policy or guardrail violations to simplify compliance."Source: [netskope.com](https://www.netskope.com/products/ai-command-center), read 2026-11-16. - **netskope-25** (vendor): Netskope says its Cloud Confidence Index covers over 370 genAI apps and 82,000+ SaaS applications.Quote: "Netskope Cloud Confidence Index (CCI) covers over 370 genAI apps and 82,000+ SaaS applications"Source: [netskope.com](https://www.netskope.com/products/securing-generative-ai), read 2026-11-16. - **netskope-26** (none-found): None found on 2026-11-16: looked for an exportable AI inventory or audit report (CSV, PDF) on the AI Command Center docs page; the Inventory page is described but no export is stated.Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-27** (observed): Netskope publishes AI Command Center documentation on docs.netskope.com that loads without sign-in.Quote: "AI Command Center (AICC) is a Netskope capability that provides comprehensive visibility, governance, and risk management for AI usage across your organization."Source: [docs.netskope.com](https://docs.netskope.com/en/ai-command-center), read 2026-11-16. - **netskope-28** (none-found): None found on 2026-11-16: looked for a published price at netskope.com/pricing, which returned a page-not-found page.Source: [netskope.com](https://www.netskope.com/pricing), read 2026-11-16. - **netskope-29** (vendor): Netskope publishes its AI Risk and Readiness Report, based on a survey of 1,253 security professionals.Quote: "Based on a survey of 1,253 cybersecurity professionals, this report explores their concerns and challenges"Source: [netskope.com](https://www.netskope.com/resources/reports-guides/ai-risk-and-readiness-report), read 2026-11-16. - **netskope-30** (third-party): SecurityWeek reported that Netskope priced its IPO at $19 per share and raised $908.2 million.Quote: "priced its IPO at $19 per share and sold 47.8 million shares, raising a total of $908.2 million"Source: [securityweek.com](https://www.securityweek.com/netskope-raises-over-908-million-in-ipo), read 2026-11-16. - **netskope-31** (third-party): SecurityWeek reported that Netskope has raised roughly $1.5 billion in total.Quote: "Netskope raised a total of roughly $1.5 billion and it was valued at $7.5 billion in 2021."Source: [securityweek.com](https://www.securityweek.com/?s=netskope), read 2026-11-16. - **netskope-32** (vendor): Netskope lists Colgate-Palmolive among its published customer stories.Quote: "Colgate-Palmolive Safeguards its “Intellectual Property” with Smart and Adaptable Data Protection"Source: [netskope.com](https://www.netskope.com/customers), read 2026-11-16. - **netskope-33** (vendor): Netskope promotes a 2026 Gartner Magic Quadrant Leader placement for SASE platforms and Security Service Edge.Quote: "See Why Gartner® Named Netskope a 2026 Magic Quadrant™ Leader for Secure Access Service Edge Platforms and Security Service Edge"Source: [netskope.com](https://www.netskope.com/company/newsroom/press), read 2026-11-16. - **netskope-34** (third-party): SiliconANGLE reported on 2026-03-11 that Netskope One AI Security provides discovery, visibility and real-time control of AI applications, models, agents and tools.Quote: "One AI Security provides comprehensive discovery, visibility and real-time control of AI applications, models, agents and tools in use"Source: [siliconangle.com](https://siliconangle.com/2026/03/11/netskope-launches-netskope-one-ai-security-suite-protect-agentic-ai-enterprise-models), read 2026-11-16. - **netskope-35** (vendor): On 2026-10-06 Netskope announced the AgentSkope Marketplace for buying, deploying and scaling AI agents.The exact quote is on the source page.Source: [netskope.com](https://www.netskope.com/company/newsroom/press), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Microsoft profile](/vendors/microsoft/) - [Nudge Security profile](/vendors/nudge-security/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Nudge Security for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/nudge-security/ Updated: 2026-11-16 Nudge Security discovers AI use without an endpoint agent or proxy. It starts with one read-only API connection to Microsoft 365 or Google Workspace email, then adds identity providers (Okta, Entra ID), SaaS APIs and an optional browser extension for Mac and Windows. It reports AI apps, accounts, OAuth integrations, remote MCP connections and agents on 23 platforms. Nudge Security says most organizations deploy in under an hour. Actions include nudging staff toward approved tools, one-click redaction and auto-revoking permissive OAuth grants on a paid tier. It publishes prices: $750 per month for Essential, $5 per user per month for Growth. None found on 16 Nov 2026 for desktop AI apps, IDE plugins or command-line agents, or hard blocking. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Layered discovery from email, identity provider, SaaS APIs and a browser extension, with a published price and a setup under an hour (nudge-security-08, nudge-security-12, nudge-security-26). **Weakness:** None found on 2026-11-16 for desktop AI apps, IDE plugins or command-line agents, and it nudges rather than blocks (nudge-security-04, nudge-security-19). ## Scores and evidence Nudge Security is 4th in the overall ranking at 64.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Nudge Security: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | AI apps, accounts, OAuth integrations, remote MCP integrations, named agent platforms and browser extensions. None found for desktop AI apps, IDE plugins or CLI agents. | [nudge-security-01](#fact-nudge-security-01), [nudge-security-02](#fact-nudge-security-02), [nudge-security-03](#fact-nudge-security-03), [nudge-security-05](#fact-nudge-security-05), [nudge-security-06](#fact-nudge-security-06), [nudge-security-04](#fact-nudge-security-04) | | Where it looks | 7 of 10 | 20 | Email, named identity providers, SaaS APIs and a browser extension; vendor-only. None found for endpoint or network sources. | [nudge-security-08](#fact-nudge-security-08), [nudge-security-11](#fact-nudge-security-11), [nudge-security-09](#fact-nudge-security-09), [nudge-security-10](#fact-nudge-security-10) | | Time to a first inventory | 7 of 10 | 15 | Vendor publishes an inventory within minutes from one read-only API connection, with OS and browsers listed. | [nudge-security-14](#fact-nudge-security-14), [nudge-security-13](#fact-nudge-security-13), [nudge-security-12](#fact-nudge-security-12), [nudge-security-09](#fact-nudge-security-09) | | What it can do about a finding | 6 of 10 | 15 | Nudges, OAuth auto-revoke, redaction and approval statuses. None found for hard blocking or ticket routing. | [nudge-security-15](#fact-nudge-security-15), [nudge-security-16](#fact-nudge-security-16), [nudge-security-17](#fact-nudge-security-17), [nudge-security-18](#fact-nudge-security-18), [nudge-security-19](#fact-nudge-security-19) | | Risk context and reporting | 6 of 10 | 10 | Training-policy summaries, risk scores and detections by user and app, with API access. None found for an exportable inventory report. | [nudge-security-20](#fact-nudge-security-20), [nudge-security-23](#fact-nudge-security-23), [nudge-security-22](#fact-nudge-security-22), [nudge-security-24](#fact-nudge-security-24), [nudge-security-25](#fact-nudge-security-25) | | Transparency | 7 of 10 | 10 | Published prices, a public changelog and public research, vendor-stated. | [nudge-security-26](#fact-nudge-security-26), [nudge-security-27](#fact-nudge-security-27), [nudge-security-28](#fact-nudge-security-28), [nudge-security-29](#fact-nudge-security-29) | | Proof | 6 of 10 | 5 | Series A reported by SecurityWeek and named customers; analyst mention is a quote on the vendor site. | [nudge-security-30](#fact-nudge-security-30), [nudge-security-32](#fact-nudge-security-32), [nudge-security-31](#fact-nudge-security-31), [nudge-security-33](#fact-nudge-security-33), [nudge-security-34](#fact-nudge-security-34) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 4th | 64.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 4th | 63.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 3rd | 65.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 4th | 64.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 4th | 64.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 4th | 63.0 | ## Product, deployment, price and owner - Product assessed: Nudge Security platform: AI and SaaS discovery, AI agent discovery, AI conversation monitoring, browser extension, AI governance nudges - Deployment: Email-provider API integration, identity provider and SaaS connected apps, plus an optional browser extension; no endpoint agent or proxy - Time to deploy: Under an hour (vendor claim); 'Deploy in 5 minutes' on homepage - Price: Essential $750/month up to 150 users; Growth $5/user/month for 150 to 1,500 users, billed annually; Enterprise custom - Owner: Nudge Security, Inc. (standalone, venture-backed; no parent company found) - Website: [nudgesecurity.com](https://www.nudgesecurity.com) ## Facts and sources - **nudge-security-01** (vendor): Nudge Security says it can discover AI apps, user accounts, OAuth and API integrations between AI tools and other apps, MCP integrations and AI agents.Quote: "AI apps, user accounts, OAuth integrations between AI tools and other apps, API integrations between AI tools and other apps, MCP integrations, AI agents"Source: [nudgesecurity.com](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. - **nudge-security-02** (vendor): Nudge Security says its API-based channel finds agents built on Salesforce Agentforce, Microsoft Copilot Studio, Gemini, ServiceNow, n8n, Tines, Notion, ChatGPT and others.Quote: "Salesforce Agentforce, Microsoft Copilot Studio, Gemini for Google Workspace, ServiceNow, n8n, Tines, Notion, ChatGPT, Abacus.AI, and Workato."Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/ai-agent-discovery), read 2026-11-16. - **nudge-security-03** (vendor): Nudge Security says its browser extension inventories browser extensions in use across the workforce to flag risky or non-compliant add-ons.Quote: "Automatically discover and inventory browser extensions in use across your workforce to identify risky or non-compliant add-ons."Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16. - **nudge-security-04** (none-found): None found on 2026-11-16: no documentation on this AI security page of discovery of desktop AI apps, IDE plugins or command-line coding agents on the endpoint; the page says discovery needs no endpoint agents.Source: [nudgesecurity.com](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. - **nudge-security-05** (vendor): Nudge Security says its browser extension detects OAuth grants that power MCP server integrations, which covers remote MCP servers connected to AI tools.Quote: "We automatically detect OAuth grants that are powering MCP server integrations using our browser extension."Source: [nudgesecurity.com](https://www.nudgesecurity.com/post/how-to-monitor-remote-mcp-server-connections-with-nudge-security), read 2026-11-16. - **nudge-security-06** (vendor): Nudge Security says it shows which SaaS vendors use third-party AI services such as OpenAI and Anthropic, which is its coverage of AI inside SaaS supply chains.Quote: "Know which of your SaaS vendors use third-party AI services like OpenAI and Anthropic as part of their service delivery or business operations."Source: [nudgesecurity.com](https://www.nudgesecurity.com/pricing), read 2026-11-16. - **nudge-security-07** (vendor): On 2026-11-02, Nudge Security said AI agent discovery across 23 agentic platforms, by API and browser extension, is available to all customers.Quote: "AI agent discovery is now available to all customers."Source: [nudgesecurity.com](https://www.nudgesecurity.com/changelog), read 2026-11-16. - **nudge-security-08** (vendor): Nudge Security says discovery starts from email analysis through a Microsoft 365 or Google Workspace integration, then adds a browser extension and SaaS API integrations.Quote: "email-based discovery that works via a lightweight integration into your organization’s email provider (Microsoft 365 or Google Workspace)"Source: [nudgesecurity.com](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. - **nudge-security-09** (vendor): Nudge Security says the browser extension deploys on Mac and Windows devices across Chromium-based browsers, Edge, Firefox and Safari, individually or through MDM.Quote: "Deploy the browser extension across Mac and Windows devices with support for Chromium-based browsers, Microsoft Edge, Firefox, Safari, and more"Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16. - **nudge-security-10** (vendor): Nudge Security says it does not touch the corporate network or endpoints, so no network, proxy or endpoint vantage point is documented.Quote: "Unlike a CASB or SASE solution, Nudge Security does not touch the corporate network or endpoints."Source: [nudgesecurity.com](https://www.nudgesecurity.com/product), read 2026-11-16. - **nudge-security-11** (vendor): Nudge Security says it integrates with Microsoft 365, Google Workspace, Okta and Entra ID, and needs no endpoint agents or network proxies.Quote: "Our platform integrates with your existing identity providers (Microsoft 365, Google Workspace, Okta, Entra ID)"Source: [nudgesecurity.com](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. - **nudge-security-12** (vendor): Nudge Security says most organizations can deploy it in under an hour.Quote: "Most organizations can deploy Nudge Security in under an hour."Source: [nudgesecurity.com](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. - **nudge-security-13** (vendor): Nudge Security says setup is one read-only API connection to a Microsoft 365 or Google Workspace email provider.Quote: "Get started with a single point of integration: read-only API access with your Microsoft 365 or Google Workspace email provider."Source: [nudgesecurity.com](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution), read 2026-11-16. - **nudge-security-14** (vendor): Nudge Security says a full AI and SaaS inventory appears within minutes of starting a free trial.Quote: "a full, continuously updated inventory of every SaaS and AI app and account within minutes of starting a free trial"Source: [nudgesecurity.com](https://www.nudgesecurity.com/comparison/nudge-security-vs-ai-security), read 2026-11-16. - **nudge-security-15** (vendor): Nudge Security says the browser extension nudges employees away from unapproved apps toward sanctioned tools and delivers the AI acceptable use policy.Quote: "Nudge Security uses your existing IT policies to automatically nudge your workforce in the browser, guiding them away from unapproved apps and toward sanctioned tools"Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16. - **nudge-security-16** (vendor): Nudge Security lists integration posture management that auto-revokes overly permissive OAuth grants (marked as a paid-tier item).Quote: "auto-revoking overly permissive and non-permitted data access through third-party OAuth grants"Source: [nudgesecurity.com](https://www.nudgesecurity.com/pricing), read 2026-11-16. - **nudge-security-17** (vendor): On 2026-11-09, Nudge Security said employees can redact detected sensitive data in one click and admins can enforce redaction in the browser.Quote: "optionally enforce redaction so unredacted data never leaves the browser."Source: [nudgesecurity.com](https://www.nudgesecurity.com/changelog), read 2026-11-16. - **nudge-security-18** (vendor): Nudge Security lets admins set an approval status of Approved, Allowed, In Review or Not Permitted for each discovered agent.Quote: "Approved, Allowed, In Review, or Not Permitted"Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/ai-agent-discovery), read 2026-11-16. - **nudge-security-19** (vendor): A customer quoted on Nudge Security's page says the product nudges instead of blocking access; no hard block of AI web apps is documented there.Quote: "Instead of blocking access, the nudge provides a helpful message, which is just great."Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16. - **nudge-security-20** (vendor): Nudge Security says it summarizes each AI vendor's model-training policy, opt-out options and retention.Quote: "understand if your data is used for model training, available opt-out options, retention periods, and other relevant information"Source: [nudgesecurity.com](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution), read 2026-11-16. - **nudge-security-21** (vendor): Nudge Security says it offers over 200,000 SaaS vendor security profiles to speed third-party risk assessments.Quote: "Access over 200,000 SaaS vendor security profiles to accelerate your third-party risk assessments."Source: [nudgesecurity.com](https://www.nudgesecurity.com/), read 2026-11-16. - **nudge-security-22** (vendor): Nudge Security says AI conversation monitoring shows detections by user, account and app, filterable by data type or action.Quote: "See conversation details and the user, account, and app level and filter based on data type or action."Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/ai-conversation-monitoring), read 2026-11-16. - **nudge-security-23** (vendor): On 2026-09-30, Nudge Security said adaptive risk management gives always-on app risk scores from 30 internal and external factors.Quote: "always-on app risk scores driven by 30 internal and external factors, with no vendor participation or manual entry required"Source: [nudgesecurity.com](https://www.nudgesecurity.com/changelog), read 2026-11-16. - **nudge-security-24** (vendor): Nudge Security says conversation data is available by API for ingestion into customer platforms and incident response.Quote: "Access conversation data via API to ingest it into your own platforms or analysis workflows for incident response."Source: [nudgesecurity.com](https://www.nudgesecurity.com/features/ai-conversation-monitoring), read 2026-11-16. - **nudge-security-25** (none-found): None found on 2026-11-16: no exportable AI inventory report or CSV export described on the pricing page.Source: [nudgesecurity.com](https://www.nudgesecurity.com/pricing), read 2026-11-16. - **nudge-security-26** (vendor): Nudge Security publishes prices: Essential $750 per month up to 150 users, Growth $5 per user per month for 150 to 1,500 users, Enterprise custom.Quote: "The full platform, priced to scale with your headcount."Source: [nudgesecurity.com](https://www.nudgesecurity.com/pricing), read 2026-11-16. - **nudge-security-27** (vendor): Nudge Security publishes a dated public changelog with an RSS feed.Quote: "to stay updated on everything we ship at Nudge Security"Source: [nudgesecurity.com](https://www.nudgesecurity.com/changelog), read 2026-11-16. - **nudge-security-28** (vendor): Nudge Security publishes public research-style blog posts on MCP connection risk, readable without sign-in.Quote: "Which MCP server connections exist in my environment, and what can they do with my data?"Source: [nudgesecurity.com](https://www.nudgesecurity.com/post/how-to-monitor-remote-mcp-server-connections-with-nudge-security), read 2026-11-16. - **nudge-security-29** (vendor): Nudge Security lists published AI adoption research on its press page.Quote: "AI Adoption Research from Nudge Security Reveals How Widespread AI Use Is Transforming Security Governance"Source: [nudgesecurity.com](https://www.nudgesecurity.com/press), read 2026-11-16. - **nudge-security-30** (third-party): SecurityWeek reported on 2025-11-18 that Nudge Security raised a $22.5 million Series A led by Cerberus Ventures.Quote: "The new investment round was led by Cerberus Ventures, with additional support from previous investors Ballistic Ventures, Forgepoint Capital, and Squadra Ventures."Source: [securityweek.com](https://www.securityweek.com/nudge-security-raises-22-5-million-in-series-a-funding/), read 2026-11-16. - **nudge-security-31** (vendor): Nudge Security's homepage quotes a Netflix director saying Netflix uses it as the record of employee apps.Quote: "Nudge is now being used by Security, Workforce Productivity, and Finance as the record of what apps employees are using."Source: [nudgesecurity.com](https://www.nudgesecurity.com/), read 2026-11-16. - **nudge-security-32** (third-party): A PR Newswire release dated 2026-09-30 quotes a Mercado Libre security engineer as a Nudge Security user.Quote: "Nudge provides a relational view across applications, users, integrations, browser extensions, and authentication that surfaces where security actually needs improvement,"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/nudge-security-unveils-adaptive-risk-management-to-track-how-saas-and-ai-risk-changes-as-usage-evolves-after-approval-302893378.html), read 2026-11-16. - **nudge-security-33** (vendor): Nudge Security's homepage quotes IDC analyst Frank Dickson on its approach.Quote: "Nudge Security innovates beyond other cloud and SaaS security technologies by providing SaaS context quickly."Source: [nudgesecurity.com](https://www.nudgesecurity.com/), read 2026-11-16. - **nudge-security-34** (vendor): Nudge Security's homepage shows a 4.7 of 5 rating on Gartner (vendor-displayed).Quote: "4.7/5 on Gartner"Source: [nudgesecurity.com](https://www.nudgesecurity.com/), read 2026-11-16. - **nudge-security-35** (third-party): SecurityWeek says Nudge Security was founded in 2021 and is based in Austin, Texas.Quote: "Founded in 2021, Austin, Texas-based"Source: [securityweek.com](https://www.securityweek.com/nudge-security-raises-22-5-million-in-series-a-funding/), read 2026-11-16. - **nudge-security-36** (vendor): Nudge Security lists Russell Spitler as co-founder and CEO and Jaime Blasco as co-founder and CTO.Quote: "Co-Founder / CEO"Source: [nudgesecurity.com](https://www.nudgesecurity.com/about-us), read 2026-11-16. - **nudge-security-37** (third-party): A PR Newswire release says Nudge Security announced AI agents (Browser Extension Risk Analyst, OAuth analysis) to select customers, with human-in-the-loop remediation.Quote: "automate remediation with human-in-the-loop decisions"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/nudge-security-unveils-ai-agents-to-mitigate-escalating-risks-from-hidden-oauth-grants-and-browser-extensions-302824148.html), read 2026-11-16. - **nudge-security-38** (vendor): On 2026-05-27, Nudge Security announced browser-based discovery of shadow AI agents on platforms without public APIs, such as Cursor Automations and Atlassian Rovo.Quote: "Cursor Automations, HyperAgent Agents, OpenAI Workflows, Retool Agents, Zapier Agents, Zoom Workflows"Source: [nudgesecurity.com](https://www.nudgesecurity.com/press/nudge-security-becomes-the-first-ai-security-solution-to-discover-shadow-ai-agents-beyond-apis), read 2026-11-16. ## News mentioning Nudge Security - [Nudge Security adds OpenAI Agents connected app](/news/nudge-security-adds-openai-agents-connected-app/) - [Nudge Security adds regional data checks for AI chats](/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/) - [Nudge Security discovers AI agents across 23 platforms](/news/nudge-security-ai-agent-discovery-across-23-platforms/) ## Related pages - [All vendor profiles](/vendors/) - [Netskope profile](/vendors/netskope/) - [Palo Alto Networks profile](/vendors/palo-alto-networks/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Palo Alto Networks for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/palo-alto-networks/ Updated: 2026-11-16 Palo Alto Networks offers AI Access Security, which reads firewall and secure access traffic, and Prisma Browser. Its documentation, readable without sign-in, describes discovery of GenAI apps against its app catalog, with a per-app risk profile and a users-per-app dashboard. Admins can sanction, coach, block data transfers inline and revoke access by privilege scope from one console. Installers are documented for Windows and macOS. None found on 16 Nov 2026 for discovery of MCP servers, AI embedded in SaaS or identity provider sources, for approval flows or ticket routing, or for a published price or time to a first inventory. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Network and browser discovery of 4000+ GenAI apps with sanction, coach, block and revoke actions in one console (palo-alto-networks-05, palo-alto-networks-18, palo-alto-networks-19). **Weakness:** None found on 2026-11-16 for MCP server, embedded SaaS AI or identity-provider discovery, or a published time to first inventory (palo-alto-networks-11, palo-alto-networks-12, palo-alto-networks-15). ## Scores and evidence Palo Alto Networks is 7th in the overall ranking at 55.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Palo Alto Networks: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 5 of 10 | 25 | GenAI apps, agents, plugins (observed in docs) and extensions. None found for MCP servers, embedded SaaS AI or desktop and coding tools. | [palo-alto-networks-05](#fact-palo-alto-networks-05), [palo-alto-networks-07](#fact-palo-alto-networks-07), [palo-alto-networks-08](#fact-palo-alto-networks-08), [palo-alto-networks-06](#fact-palo-alto-networks-06), [palo-alto-networks-11](#fact-palo-alto-networks-11) | | Where it looks | 6 of 10 | 20 | Docs observe firewall, browser access and SaaS posture sources. None found for an identity provider or endpoint agent for discovery. | [palo-alto-networks-06](#fact-palo-alto-networks-06), [palo-alto-networks-10](#fact-palo-alto-networks-10), [palo-alto-networks-07](#fact-palo-alto-networks-07), [palo-alto-networks-12](#fact-palo-alto-networks-12), [palo-alto-networks-13](#fact-palo-alto-networks-13) | | Time to a first inventory | 5 of 10 | 15 | Deploys through existing gateways with no agent; docs observe Windows and macOS installers. None found for a time to first inventory. | [palo-alto-networks-14](#fact-palo-alto-networks-14), [palo-alto-networks-16](#fact-palo-alto-networks-16), [palo-alto-networks-17](#fact-palo-alto-networks-17), [palo-alto-networks-15](#fact-palo-alto-networks-15) | | What it can do about a finding | 6 of 10 | 15 | Revoke access by privilege scope, coaching and inline blocking of data transfers. None found for approval flows or tickets. | [palo-alto-networks-18](#fact-palo-alto-networks-18), [palo-alto-networks-19](#fact-palo-alto-networks-19), [palo-alto-networks-20](#fact-palo-alto-networks-20) | | Risk context and reporting | 5 of 10 | 10 | Per-app risk profile and observed users-per-app dashboard. None found for an export or a training-on-inputs rating. | [palo-alto-networks-21](#fact-palo-alto-networks-21), [palo-alto-networks-22](#fact-palo-alto-networks-22), [palo-alto-networks-23](#fact-palo-alto-networks-23) | | Transparency | 6 of 10 | 10 | Documentation observed readable without sign-in and a research report. None found for a price. | [palo-alto-networks-25](#fact-palo-alto-networks-25), [palo-alto-networks-26](#fact-palo-alto-networks-26), [palo-alto-networks-24](#fact-palo-alto-networks-24) | | Proof | 7 of 10 | 5 | NASDAQ-listed with 70,000+ customers, a Gartner recognition and a named customer, all vendor-stated; The Register covers the product line, not customers, funding or analysts. | [palo-alto-networks-28](#fact-palo-alto-networks-28), [palo-alto-networks-27](#fact-palo-alto-networks-27), [palo-alto-networks-31](#fact-palo-alto-networks-31), [palo-alto-networks-70](#fact-palo-alto-networks-70) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 7th | 55.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 8th | 55.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 8th | 54.0 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 8th | 55.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | joint 7th | 56.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 9th | 56.5 | ## Product, deployment, price and owner - Product assessed: AI Access Security, with Prisma Browser - Deployment: AI Access Security via existing NGFW and SSE gateways with no endpoint agent; Prisma Browser as a browser, extension or mobile app; Prisma AIRS via API Intercept, Network Intercept and AI Gateway (SaaS or Hybrid) - Time to deploy: none found on 2026-11-16 (Prisma Browser: access 'in minutes' per vendor) - Price: none found on 2026-11-16 - Owner: Palo Alto Networks, Inc. (NASDAQ: PANW) - Website: paloaltonetworks.com ## Facts and sources - **palo-alto-networks-01** (vendor): The Palo Alto Networks leadership page lists Nikesh Arora as chairman and CEO, so the company is its own parent and the owner of Prisma AIRS, AI Access Security and Prisma Browser.Quote: "Nikesh Arora is chairman and CEO of Palo Alto Networks, the world’s leading cybersecurity company."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/about-us/management), read 2026-11-16. - **palo-alto-networks-02** (vendor): On 2026-05-29 Palo Alto Networks said it closed the acquisition of Portkey, an AI gateway company, and made it the AI Gateway of Prisma AIRS.Quote: "today announced it has closed the acquisition of Portkey, a pioneer in AI Gateways."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/company/press/2026/palo-alto-networks-completes-acquisition-of-portkey-to-secure-ai-agents), read 2026-11-16. - **palo-alto-networks-03** (vendor): Palo Alto Networks documents Prisma AIRS as its AI security platform covering AI applications, models, agents and infrastructure across the AI lifecycle.Quote: "Prisma® AIRS™ is Palo Alto Networks' comprehensive AI security platform."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-runtime-security), read 2026-11-16. - **palo-alto-networks-04** (none-found): None found on 2026-11-16: endpoint-agent coverage for desktop or command-line agents inside the Prisma AIRS product page; the page lists Agentic Endpoint Security as a separate capability, which is a separate profile and not counted here.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/prisma/prisma-ai-runtime-security), read 2026-11-16. - **palo-alto-networks-05** (vendor): Palo Alto Networks says AI Access Security uses a GenAI app dictionary to discover and categorize GenAI applications, agents and marketplace plugins.Quote: "Access the most up-to-date GenAI app dictionary to accurately discover and categorize GenAI applications, agents and marketplace plugins."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-06** (vendor): Palo Alto Networks says AI Access Security discovers AI tools by continuously monitoring network traffic.Quote: "AI access security continuously monitors network traffic to discover and categorize all AI tools used across an organization."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-07** (observed): The AI Access Security docs describe a dashboard of GenAI apps installed as third-party plugins, with plugin and user counts drawn from SaaS Security Posture Management.Quote: "Discover the risks posed by GenAI apps that users installed as third-party plugins."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps), read 2026-11-16. - **palo-alto-networks-08** (vendor): Palo Alto Networks says Prisma Browser discovers all extensions in use, monitors them and blocks risky or over-permissioned ones.Quote: "Discover all extensions in use, continuously monitor them for threats, and block risky or over-permissioned extensions."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/prisma-browser), read 2026-11-16. - **palo-alto-networks-09** (vendor): Palo Alto Networks claims Prisma Browser gives visibility into all GenAI apps and agentic workflows.Quote: "Visibility into all GenAI apps and agentic workflows, effectively eliminating all shadow AI."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/prisma-browser), read 2026-11-16. - **palo-alto-networks-10** (observed): The AI Access Security docs list the places it reads from: NGFW, Prisma Access and Prisma Browser, with Strata Logging Service.Quote: "NGFW (Managed by Panorama or Strata Cloud Manager) Prisma Access (Managed by Panorama or Strata Cloud Manager) Prisma Browser"Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps), read 2026-11-16. - **palo-alto-networks-11** (none-found): None found on 2026-11-16: discovery of MCP servers or AI features embedded in approved SaaS apps on the AI Access Security product page and its docs home page.Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security), read 2026-11-16. - **palo-alto-networks-12** (none-found): None found on 2026-11-16: an identity-provider integration (Okta, Entra ID, Google) used for AI app discovery on the AI Access Security product page and docs home page.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-13** (none-found): None found on 2026-11-16: an endpoint agent used for AI app discovery in AI Access Security; the vendor says it needs no endpoint agents, and endpoint governance sits in the separate Agentic Endpoint Security profile.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-14** (vendor): Palo Alto Networks says AI Access Security deploys through existing SSE and firewall gateways with no additional endpoint agents or infrastructure changes.The exact quote is on the source page.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-15** (none-found): None found on 2026-11-16: a published time to a first AI inventory for AI Access Security on its product page or docs home page.Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security), read 2026-11-16. - **palo-alto-networks-16** (vendor): Palo Alto Networks documents Prisma Browser as a browser, an extension for consumer browsers and a mobile app, with access in minutes.Quote: "Available as a browser, an extension for consumer browsers, and a mobile app, it provides secure access from any location in minutes."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/prisma-browser), read 2026-11-16. - **palo-alto-networks-17** (observed): The Prisma Browser deployment docs describe Windows MSI installers for x64 and ARM64 and a macOS installation, with weekly updates.Quote: "Prisma Browser supports two installation variations for macOS environments."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/content/techdocs/en_US/prisma-browser/deployment), read 2026-11-16. - **palo-alto-networks-18** (vendor): Palo Alto Networks says AI Access Security administrators can revoke access by privilege scope and risk and control actions such as upload and download.Quote: "Administrators can revoke access based on the scope of privileges and risk factors, as well as apply fine-grained control over actions such as upload and download."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-19** (vendor): Palo Alto Networks says AI Access Security shows notifications and coaching when employees try unsanctioned GenAI apps or may violate AI usage policy.Quote: "Employees can understand and correct risky behavior when attempting to access unsanctioned GenAI apps or are about to violate AI usage policies."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-20** (vendor): Palo Alto Networks says AI Access Security inline detection blocks sensitive text and file transfers to GenAI apps.Quote: "Inline data detection ensures regulatory compliance and blocks sensitive text- and file-based data transfer to GenAI apps."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-21** (vendor): Palo Alto Networks says AI Access Security assesses each app's risk profile so teams can allow read-only access or block high-risk tools.Quote: "It assesses application risk profiles, allowing security teams to enforce policies like granting read-only access or blocking high-risk tools."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-22** (observed): The AI Access Security docs describe an Insights dashboard showing which GenAI apps are used and by which users.Quote: "The AI Access Security Insights dashboard provides in-depth details to help you understand which GenAI apps are being used and by who."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps), read 2026-11-16. - **palo-alto-networks-23** (none-found): None found on 2026-11-16: a documented exportable AI inventory or audit report, or a per-app data-use and training-on-inputs rating, on the AI Access Security product page.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-24** (none-found): None found on 2026-11-16: a published price for Prisma AIRS, AI Access Security or Prisma Browser on their product pages; guessed pricing URLs on paloaltonetworks.com returned 404.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-25** (observed): The AI Access Security documentation is readable without sign-in and describes onboarding and policy configuration.Quote: "AI Access Security enables organizations to safely adopt GenAI applications by employees by mitigating the risks posed by inadvertent data leakage in prompts and malicious content in responses."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/ai-access-security), read 2026-11-16. - **palo-alto-networks-26** (vendor): Palo Alto Networks lists a published research report on prompt attacks among the AI Access Security resources.Quote: "Securing GenAI: A Comprehensive Report on Prompt Attacks: Taxonomy, Risks and Solutions"Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. - **palo-alto-networks-27** (vendor): Palo Alto Networks says Gartner recognized it as a Market Shaper in the Emerging Market Quadrant for AI Application Security in September 2026.The exact quote is on the source page.Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/prisma/prisma-ai-runtime-security), read 2026-11-16. - **palo-alto-networks-28** (vendor): Palo Alto Networks says it is a NASDAQ-listed company with more than 70,000 customers.Quote: "Trusted by 70,000+ customers and powered by Unit 42 threat intelligence"Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/company/press/2026/palo-alto-networks-completes-acquisition-of-portkey-to-secure-ai-agents), read 2026-11-16. - **palo-alto-networks-31** (vendor): The Prisma Browser page quotes Wipro's CISO as a named customer.Quote: "Prisma Browser delivered true zero trust security to the last mile."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/sase/prisma-browser), read 2026-11-16. - **palo-alto-networks-44** (vendor): The AI Gateway docs offer a SaaS model and a Hybrid model where the data plane runs in the customer's own Kubernetes cluster.Quote: "Hybrid , where you host the data plane in your own Kubernetes cluster so that request and response payloads never leave your environment."Source: [docs.paloaltonetworks.com](https://docs.paloaltonetworks.com/prisma-airs/ai-gateway/ai-gateway-overview), read 2026-11-16. - **palo-alto-networks-52** (vendor): Palo Alto Networks says Prisma AIRS offers Network Intercept and API Intercept deployment modes.Quote: "It offers deployment modes such as Network Intercept and API Intercept to secure AI applications, AI models, AI data, and AI agents."Source: [paloaltonetworks.com](https://www.paloaltonetworks.com/prisma/prisma-ai-runtime-security), read 2026-11-16. - **palo-alto-networks-70** (third-party): On 2025-08-19 The Register reported that Palo Alto Networks owns the Prisma Access Browser (now Prisma Browser) and sells it with its SASE products; this is older than the 90-day window.The exact quote is on the source page.Source: [theregister.com](https://www.theregister.com/security/2025/08/19/browser-wars-are-back-predicts-palo-alto-thanks-to-ai/1228881), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Nudge Security profile](/vendors/nudge-security/) - [Push Security profile](/vendors/push-security/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Push Security for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/push-security/ Updated: 2026-11-16 Push Security is a browser extension, deployed through Intune, Group Policy, Google Admin, macOS management or self-enrollment, that watches logins. It discovers AI apps, AI browser extensions, AI OAuth grants and agentic browsers, and can monitor, warn with alternatives, block or redact on paste, with Jira tickets through a webhook. It publishes a price: Standard is $5 per user per month for up to 500 employees. Self-enrollment takes about a minute per user. The vantage point is the browser only. None found on 16 Nov 2026 for endpoint, network or SaaS log sources, a per-tool AI risk score, or a time to a first inventory. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Discovers AI apps, AI browser extensions and AI OAuth grants from browser logins and can monitor, warn or block, with a published $5 per user per month price (push-security-03, push-security-04, push-security-05, push-security-17, push-security-26). **Weakness:** Browser-only vantage: no endpoint, network or SaaS-log source and no per-tool AI risk rating or time to first inventory; none found on 2026-11-16 (push-security-16, push-security-21, push-security-25). ## Scores and evidence Push Security is joint 9th in the overall ranking at 53.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Push Security: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 5 of 10 | 25 | AI apps from browser logins, AI extensions, OAuth connections and agentic browsers are claimed. None found for desktop AI apps, IDE plugins, CLI agents or embedded SaaS AI. | [push-security-03](#fact-push-security-03), [push-security-04](#fact-push-security-04), [push-security-05](#fact-push-security-05), [push-security-06](#fact-push-security-06), [push-security-07](#fact-push-security-07), [push-security-08](#fact-push-security-08) | | Where it looks | 4 of 10 | 20 | Browser extension with identity provider sync and OAuth visibility into Google Workspace. None found for endpoint, network, email or billing sources. | [push-security-09](#fact-push-security-09), [push-security-10](#fact-push-security-10), [push-security-05](#fact-push-security-05), [push-security-12](#fact-push-security-12) | | Time to a first inventory | 6 of 10 | 15 | Silent extension install via MDM or one-minute self-enrollment, desktop-only per TechCrunch. None found for a time to first inventory. | [push-security-13](#fact-push-security-13), [push-security-14](#fact-push-security-14), [push-security-15](#fact-push-security-15), [push-security-12](#fact-push-security-12), [push-security-16](#fact-push-security-16) | | What it can do about a finding | 7 of 10 | 15 | Block, warn with alternatives, redact on paste and Jira tickets via webhook. None found for OAuth revocation. | [push-security-17](#fact-push-security-17), [push-security-18](#fact-push-security-18), [push-security-19](#fact-push-security-19), [push-security-20](#fact-push-security-20), [push-security-21](#fact-push-security-21) | | Risk context and reporting | 4 of 10 | 10 | Per-tool sign-in and MFA data, categories and API extraction. None found for a training-on-inputs or vendor posture rating. | [push-security-22](#fact-push-security-22), [push-security-23](#fact-push-security-23), [push-security-24](#fact-push-security-24), [push-security-25](#fact-push-security-25) | | Transparency | 7 of 10 | 10 | Published price, public documentation and published research, all vendor-stated. | [push-security-26](#fact-push-security-26), [push-security-27](#fact-push-security-27), [push-security-28](#fact-push-security-28) | | Proof | 5 of 10 | 5 | Series B claimed, 1.5 million endpoints and early customers reported by third parties. None found for analyst coverage. | [push-security-29](#fact-push-security-29), [push-security-30](#fact-push-security-30), [push-security-31](#fact-push-security-31), [push-security-32](#fact-push-security-32), [push-security-34](#fact-push-security-34) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | joint 9th | 53.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 12th | 50.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 9th | 52.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 10th | 52.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 11th | 51.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 9th | 56.5 | ## Product, deployment, price and owner - Product assessed: Push browser security platform: AI visibility and control, shadow SaaS and app discovery, browser extension and OAuth inventory - Deployment: Browser extension deployed through device management software (Intune, Group Policy, Google Admin, macOS MDM, Island) or self-enrollment - Time to deploy: none found on 2026-11-16 (self-enrollment per user takes about a minute) - Price: Standard $5 per user per month for up to 500 employees; Enterprise (500+) by quote - Owner: Push Security (company named 'Push' in its 2026 site footer) - Website: [pushsecurity.com](https://pushsecurity.com) ## Facts and sources - **push-security-01** (vendor): Push Security describes its product as a Secure Enterprise Browser extension that covers attacks, AI security, identity hardening and data loss prevention.Quote: "The Secure Enterprise Browser extension that combines browser telemetry, real-time control, and autonomous agents to stop attacks, secure AI, harden identities, and prevent data loss."Source: [pushsecurity.com](https://pushsecurity.com), read 2026-11-16. - **push-security-02** (third-party): SiliconANGLE reports that Push Security was founded in 2020 and sells a browser-based identity threat detection and response platform.Quote: "Founded in 2020, Push Security specializes in preventing identity-based attacks through a browser-based Identity Threat Detection and Response platform."Source: [siliconangle.com](https://siliconangle.com/2025/04/24/push-security-raises-30m-expand-browser-based-identity-threat-detection/), read 2026-11-16. - **push-security-03** (vendor): Push says it discovers and inventories AI apps from browser login events.Quote: "Push automatically discovers and inventories AI apps from browser login events."Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-04** (vendor): Push says it catalogs AI browser extensions across the workforce with the permissions each requested.Quote: "catalogs every AI-related extension installed across the workforce, including the specific permissions each extension has requested"Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-05** (vendor): Push says it identifies OAuth connections between AI tools and corporate systems such as Google Workspace.Quote: "identifies OAuth connections between AI tools and corporate systems"Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-06** (vendor): Push says it detects employees adopting agentic AI browsers.The exact quote is on the source page.Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-07** (vendor): Push says custom detections can block unapproved MCP server connections made through the browser; this is browser-layer control, not discovery of MCP servers configured on machines.Quote: "blocking unapproved MCP server connections"Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-08** (vendor): Push says download blocking exists because desktop versions of AI tools move usage outside the browser where Push has visibility.Quote: "Employees downloading desktop versions of AI tools, which moves usage outside the browser where Push has visibility."Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-09** (vendor): Push says it deploys as a lightweight extension in existing browsers and sees browser-layer activity that network and endpoint tools lack.Quote: "Push deploys as a lightweight extension to employees' existing browsers rather than requiring a full browser migration"Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-10** (vendor): Push documents integration with Microsoft 365, Google Workspace or Okta as the identity provider, used to sync employee records.Quote: "Push also integrates with your identity provider (M365, Google Workspace, or Okta)."Source: [pushsecurity.com](https://pushsecurity.com/help/audience/administrators/docs/install-the-browser-extension), read 2026-11-16. - **push-security-11** (vendor): Push says it forwards AI-related telemetry, including permitted events, to SIEMs such as Microsoft Sentinel, Splunk Cloud, Datadog, Panther and Cribl Cloud, or any SIEM reachable by webhook.Quote: "forwards it as structured, context-enriched data to Microsoft Sentinel, Splunk Cloud, Datadog, Panther, Cribl Cloud, or any SIEM reachable by webhook."Source: [pushsecurity.com](https://pushsecurity.com/news/push-ai-visibility-and-control-capabilities), read 2026-11-16. - **push-security-12** (third-party): TechCrunch reported in 2023 that Push is designed to work on desktops and laptops, not mobile.The exact quote is on the source page.Source: [techcrunch.com](https://techcrunch.com/2023/04/03/push-security-raises-15m-to-help-saas-users-lower-their-online-vulnerability/), read 2026-11-16. - **push-security-13** (vendor): Push documents that the extension is installed silently through device management software, with the managed method recommended.Quote: "The Push browser extension is deployed using device management software. This allows you to silently install the extension on all browsers in your environment"Source: [pushsecurity.com](https://pushsecurity.com/help/audience/administrators/docs/install-the-browser-extension), read 2026-11-16. - **push-security-14** (vendor): Push says per-employee self-enrollment by email takes about a minute.Quote: "This process takes about a minute."Source: [pushsecurity.com](https://pushsecurity.com/help/audience/administrators/docs/install-the-browser-extension), read 2026-11-16. - **push-security-15** (vendor): Push documents that clipboard blocking works on all supported browsers except Safari.Quote: "This feature is available on all browsers that Push supports, except Safari."Source: [pushsecurity.com](https://pushsecurity.com/help/can-push-block-clipboard-actions), read 2026-11-16. - **push-security-16** (none-found): None found on 2026-11-16: a published time to a first AI inventory after setup, looked for on the install guide.Source: [pushsecurity.com](https://pushsecurity.com/help/audience/administrators/docs/install-the-browser-extension), read 2026-11-16. - **push-security-17** (vendor): Push says block mode stops logins to unapproved apps with a blocking banner.The exact quote is on the source page.Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-18** (vendor): Push says acknowledge mode warns users and points them to approved alternatives instead of blocking.The exact quote is on the source page.Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-19** (vendor): Push says clipboard warn mode can offer a redacted version of sensitive data so the user can continue.Quote: "Push offers a redacted version of the sensitive data so the employee can continue their work"Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-20** (vendor): Push documents a REST API and webhooks that can create tickets in Jira or similar systems.Quote: "Creating a ticket in JIRA or another ticketing system to track an event that needs to be resolved."Source: [pushsecurity.com](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system), read 2026-11-16. - **push-security-21** (none-found): None found on 2026-11-16: a documented action to revoke an OAuth grant held by an AI app, looked for on the shadow AI guide and the pricing feature list.Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-22** (vendor): Push says each discovered AI tool is recorded with sign-in method, MFA status and password strength.Quote: "SSO vs. password vs. personal account, MFA status, and password strength."Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-23** (vendor): Push documents automatic app categories such as AI / ML apps, which can scope control rules.Quote: "Push will assign categories to apps it inventories based on the app’s purpose, e.g. AI / ML apps, file-sharing apps"Source: [pushsecurity.com](https://pushsecurity.com/help/does-push-automatically-categorize-apps-it-discovers), read 2026-11-16. - **push-security-24** (vendor): Push documents a REST API that can extract data for reporting and analysis.Quote: "Extract data for reporting and analysis."Source: [pushsecurity.com](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system), read 2026-11-16. - **push-security-25** (none-found): None found on 2026-11-16: a per-tool AI risk rating covering training on inputs or vendor posture, looked for on the shadow AI guide and the categorization help page.Source: [pushsecurity.com](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. - **push-security-26** (vendor): Push publishes a Standard plan price of $5 per user per month for up to 500 employees, with Enterprise on request.Quote: "Standard Up to 500 employees $ 5 User/month"Source: [pushsecurity.com](https://pushsecurity.com/pricing), read 2026-11-16. - **push-security-27** (vendor): Push publishes admin documentation and help articles that need no sign-in.Quote: "Push provides a REST API to programmatically administer the platform, as well as webhooks to support real-time updates for platform events."Source: [pushsecurity.com](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system), read 2026-11-16. - **push-security-28** (vendor): Push publishes its own telemetry research on shadow AI in April 2026.Quote: "Push telemetry shows that the average organization has 16 unique AI apps, 17 AI browser extensions, and 17 AI OAuth integrations in active use"Source: [pushsecurity.com](https://pushsecurity.com/blog/what-push-data-reveals-about-the-state-of-shadow-ai), read 2026-11-16. - **push-security-29** (vendor): Push announced a $30 million Series B led by Redpoint Ventures with Datadog Ventures and B3 Capital.Quote: "today announced a $30 million Series B funding round led by Redpoint Ventures, with participation from Datadog Ventures and B3 Capital."Source: [pushsecurity.com](https://pushsecurity.com/news/push-security-secures-30-million-series-b-funding), read 2026-11-16. - **push-security-30** (third-party): SiliconANGLE reports Push says its platform is deployed on more than 1.5 million endpoints.Quote: "its platform is now deployed on more than 1.5 million endpoints globally"Source: [siliconangle.com](https://siliconangle.com/2025/04/24/push-security-raises-30m-expand-browser-based-identity-threat-detection/), read 2026-11-16. - **push-security-31** (third-party): TechCrunch named early Push customers in 2023.Quote: "customers including Reachdesk, Upvest and Tray.io"Source: [techcrunch.com](https://techcrunch.com/2023/04/03/push-security-raises-15m-to-help-saas-users-lower-their-online-vulnerability/), read 2026-11-16. - **push-security-32** (vendor): Push quotes Myke Lyons, CISO of Cribl, as a customer on its home page.Quote: "I'm not going to be able to shove a browser on everybody."Source: [pushsecurity.com](https://pushsecurity.com), read 2026-11-16. - **push-security-33** (vendor): On 2026-09-29, Push announced a partnership to power Proofpoint Advanced Browser Protection, expected in early 2027.Quote: "Push Security today announced a partnership with Proofpoint to power new investigation, detection and response capabilities in Proofpoint Advanced Browser Protection."Source: [pushsecurity.com](https://pushsecurity.com/news/push-partners-with-proofpoint), read 2026-11-16. - **push-security-34** (none-found): None found on 2026-11-16: independent analyst coverage (Gartner, Forrester, IDC) of Push, looked for on the newsroom page.Source: [pushsecurity.com](https://pushsecurity.com/news), read 2026-11-16. - **push-security-35** (vendor): On 2026-06-02, Push launched file, clipboard, app and domain controls for unsanctioned AI tools, with an 89-category domain framework.Quote: "An 89-category framework spanning personal file sharing, unapproved AI tools, adult content, gambling, and more"Source: [pushsecurity.com](https://pushsecurity.com/news/push-ai-visibility-and-control-capabilities), read 2026-11-16. - **push-security-36** (vendor): Push's pricing page footer names the company as Push with a 2026 copyright; no separate parent company was found.Quote: "© Push 2026. All rights reserved."Source: [pushsecurity.com](https://pushsecurity.com/about), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Palo Alto Networks profile](/vendors/palo-alto-networks/) - [Reco profile](/vendors/reco/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Reco for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/reco/ Updated: 2026-11-16 Reco finds shadow AI from SSO, OAuth and email metadata, with browser behavior and network traffic also claimed. Its pages say it finds ChatGPT, Claude, AI agents, OAuth-connected AI apps, browser extensions and AI embedded in named SaaS, with a risk label and users per tool. Reco says first discovery takes minutes and full deployment 24 to 48 hours through API connections. The Hacker News describes the product as detection and alerting that cannot prevent AI tools from connecting. None found on 16 Nov 2026 for endpoint, desktop app or MCP server discovery, a published price, public documentation or an operating system list. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Finds ChatGPT, Claude, AI agents, OAuth-connected AI apps, browser extensions and embedded SaaS AI from SSO, OAuth and email metadata, with a risk label per tool (reco-02, reco-05, reco-14). **Weakness:** Described as visibility and detection that cannot prevent AI tools from connecting, with no published price, public docs, OS list or endpoint, desktop or MCP discovery found on 2026-11-16 (reco-09, reco-10, reco-18, reco-19). ## Scores and evidence Reco is joint 11th in the overall ranking at 49.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Reco: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 5 of 10 | 25 | Shadow AI tools, agents, OAuth apps, browser extensions and named embedded SaaS AI are claimed. None found for desktop apps, IDE plugins, CLI agents or MCP servers. | [reco-02](#fact-reco-02), [reco-03](#fact-reco-03), [reco-04](#fact-reco-04), [reco-06](#fact-reco-06), [reco-05](#fact-reco-05) | | Where it looks | 7 of 10 | 20 | Browser behavior, OAuth, email metadata, SSO logs and network traffic are claimed, with business-email inventory confirmed by The Hacker News. None found for endpoint or billing sources. | [reco-05](#fact-reco-05), [reco-06](#fact-reco-06), [reco-07](#fact-reco-07), [reco-09](#fact-reco-09) | | Time to a first inventory | 6 of 10 | 15 | Vendor publishes discovery in minutes and full deployment in 24 to 48 hours via API connections. None found for OS or browser coverage. | [reco-08](#fact-reco-08), [reco-06](#fact-reco-06), [reco-09](#fact-reco-09) | | What it can do about a finding | 3 of 10 | 15 | Ticketing and compliance workflows, but The Hacker News describes it as detection and alerting that cannot prevent connections. | [reco-10](#fact-reco-10), [reco-11](#fact-reco-11), [reco-12](#fact-reco-12) | | Risk context and reporting | 5 of 10 | 10 | Risk classification, vendor details and users per tool are claimed. None found for an exportable inventory or report. | [reco-13](#fact-reco-13), [reco-14](#fact-reco-14), [reco-15](#fact-reco-15), [reco-16](#fact-reco-16), [reco-17](#fact-reco-17) | | Transparency | 1 of 10 | 10 | Product pages are readable and statistics are published without method. None found for documentation, a price or a reachable report. | [reco-20](#fact-reco-20), [reco-18](#fact-reco-18), [reco-19](#fact-reco-19) | | Proof | 7 of 10 | 5 | TechCrunch and SiliconANGLE report $140 million raised, 100+ customers and a named customer investor. None found for analyst reports. | [reco-21](#fact-reco-21), [reco-22](#fact-reco-22), [reco-23](#fact-reco-23), [reco-24](#fact-reco-24) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | joint 11th | 49.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 11th | 50.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | joint 6th | 55.0 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 10th | 52.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 9th | 53.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 15th | 46.0 | ## Product, deployment, price and owner - Product assessed: Reco Shadow AI Discovery / Shadow Agent Discovery, Application Discovery and AI Governance and Security - Deployment: SaaS platform connected through APIs, SSO, OAuth and email metadata; about page says deployment in 24 hours, home page says 48h to full deployment - Time to deploy: 24 to 48 hours to full deployment as published; first discovery in minutes (vendor claim) - Price: none found on 2026-11-16 - Owner: Reco (Recolabs Inc.), standalone venture-backed company; no parent company named on its pages - Website: reco.ai ## Facts and sources - **reco-01** (vendor): Reco says it is the leader in agentic ecosystem security; the about page lists co-founders Ofer Klein (CEO), Gal Nakash (CPO) and Dr. Tal Shapira (CTO), with offices in Altamonte Springs, Florida and Tel Aviv, and names no parent company.Quote: "Reco is the leader in Agentic Ecosystem Security, securing everything around the agent, not just the prompt."Source: [reco.ai](https://www.reco.ai/about-us), read 2026-11-16. - **reco-02** (vendor): Reco says its Shadow AI Discovery tracks shadow AI tools, unauthorized GenAI use and agents outside IT oversight, including their data access patterns.Quote: "Instantly track all shadow AI tools, unauthorized GenAI usage, and agents operating outside IT oversight, including their data access patterns."Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-03** (vendor): Reco says Application Discovery monitors OAuth apps, browser extensions and APIs, so OAuth-granted AI apps and AI browser extensions are in scope.Quote: "Monitors OAuth apps, browser extensions, and APIs"Source: [reco.ai](https://www.reco.ai/platform/application-discovery), read 2026-11-16. - **reco-04** (vendor): Reco says its AI governance covers AI features embedded in existing SaaS applications, such as Salesforce Einstein and Microsoft Copilot.Quote: "From Salesforce Einstein to Microsoft Copilot, govern AI features embedded within your existing SaaS applications."Source: [reco.ai](https://www.reco.ai/platform/ai-governance-security), read 2026-11-16. - **reco-05** (vendor): Reco lists its shadow AI detection methods: browser behavior, OAuth connections, email metadata, network traffic analysis and browser extension detection.Quote: "Detects when unapproved AI applications connect to corporate systems via OAuth integrations"Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-06** (vendor): Reco says Application Discovery draws on SSO logs, email metadata and OAuth integration monitoring, and finds apps like ChatGPT and Claude without accessing private content.Quote: "SSO logs and email metadata"Source: [reco.ai](https://www.reco.ai/platform/application-discovery), read 2026-11-16. - **reco-07** (third-party): The Hacker News product review says Reco inventories applications tied to business email, shows how users authenticate, and sends alerts to SIEM or SOAR; it also notes personal ChatGPT use outside corporate infrastructure is hard to see.Quote: "It integrates with your SIEM or SOAR so organizations can remediate SaaS risks efficiently within existing workflows."Source: [thehackernews.com](https://thehackernews.com/2025/01/product-review-how-reco-discovers.html), read 2026-11-16. - **reco-08** (vendor): Reco says Application Discovery reveals a complete SaaS ecosystem in minutes, and its about page says deployment takes 24 hours (its home page shows 48h to full deployment).Quote: "Discover your complete SaaS ecosystem in minutes."Source: [reco.ai](https://www.reco.ai/platform/application-discovery), read 2026-11-16. - **reco-09** (none-found): None found on 2026-11-16: supported operating systems, browsers, or whether an endpoint agent or browser extension is required, on the shadow AI discovery, application discovery and AI governance pages; the described method is API, SSO, OAuth and email metadata.Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-10** (third-party): The Hacker News review says Reco cannot prevent third-party shadow AI tools from connecting through APIs but can identify and alert on the connections, calling it a visibility and detection tool.Quote: "Reco cannot prevent third-party shadow AI tools from connecting via APIs, but it can identify and alert these connections."Source: [thehackernews.com](https://thehackernews.com/2025/01/product-review-how-reco-discovers.html), read 2026-11-16. - **reco-11** (vendor): Reco says its SaaS ticketing workflow creates and tracks remediation tasks in the customer's existing ticketing system.The exact quote is on the source page.Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-12** (vendor): Reco says each discovered AI tool can be tagged, assigned governance, or sent into a compliance workflow.Quote: "Tag it, assign governance, or trigger a compliance workflow instantly."Source: [reco.ai](https://www.reco.ai/platform/ai-governance-security), read 2026-11-16. - **reco-13** (vendor): Reco says it classifies shadow AI tools by business risk and data exposure.Quote: "Automatically categorizes shadow AI tools by business risk and data exposure"Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-14** (vendor): Reco says every discovered app and AI agent is labeled with business context, vendor details and a security risk assessment.Quote: "Automatically label every discovered app and AI agent with business context, vendor details, and security risk assessments."Source: [reco.ai](https://www.reco.ai/platform/application-discovery), read 2026-11-16. - **reco-15** (vendor): Reco says each AI tool is tied to users, roles and data access behavior, so security teams can see who uses AI.Quote: "Every AI tool is tied to users, roles, and data access behavior."Source: [reco.ai](https://www.reco.ai/platform/ai-governance-security), read 2026-11-16. - **reco-16** (third-party): On 2026-10-22, Reco announced a Cyera integration that feeds Cyera data classification labels into the Reco Graph to show which AI agents can reach sensitive data.Quote: "Reco ingests Cyera's data classification labels, including business IP, financial, health and regulated records, and feeds them into the Reco Graph."Source: [markets.businessinsider.com](https://markets.businessinsider.com/news/stocks/reco-integrates-with-cyera-to-prioritize-ai-agent-risk-across-sensitive-data-and-applications-1036654651), read 2026-11-16. - **reco-17** (none-found): None found on 2026-11-16: a documented exportable AI inventory or audit report on the shadow AI discovery, application discovery and AI governance pages.Source: [reco.ai](https://www.reco.ai/platform/application-discovery), read 2026-11-16. - **reco-18** (none-found): None found on 2026-11-16: a published price; www.reco.ai/pricing returns a 404 page and the site routes buyers to a demo.Source: [reco.ai](https://www.reco.ai/pricing), read 2026-11-16. - **reco-19** (none-found): None found on 2026-11-16: public product documentation; docs.reco.ai did not load, and the State of Shadow AI report page (www.reco.ai/state-of-shadow-ai) returned 404.Source: [docs.reco.ai](https://docs.reco.ai), read 2026-11-16. - **reco-20** (vendor): Reco's shadow AI page publishes statistics, such as a claim that 72% of enterprise GenAI usage is through unauthorized tools, without a method on the page; the product pages are readable without sign-in.Quote: "72% of enterprise GenAI usage happens through unauthorized AI tools connected to corporate systems and data."Source: [reco.ai](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16. - **reco-21** (third-party): TechCrunch reported on 2026-09-29 that Reco raised $55 million, after a $30 million Series B in February, bringing total capital raised to $140 million, with AT&T (a customer) investing through its venture arm.Quote: "The new funding brings Reco’s total capital raised to $140 million."Source: [techcrunch.com](https://techcrunch.com/2026/09/29/reco-raises-55m-as-ai-agent-security-startups-crowd-the-market/), read 2026-11-16. - **reco-22** (third-party): TechCrunch reported Reco has more than 100 customers, with financial services about 40% of the business.Quote: "The startup has more than 100 customers, and financial services companies account for about 40% of the business."Source: [techcrunch.com](https://techcrunch.com/2026/09/29/reco-raises-55m-as-ai-agent-security-startups-crowd-the-market/), read 2026-11-16. - **reco-23** (third-party): SiliconANGLE reported on 2025-04-22 that Reco closed a $25 million round with Insight Partners, Zeev Ventures, boldstart ventures, Angular Ventures and Redseed participating.Quote: "It included the participation of Insight Partners, Zeev Ventures, boldstart ventures, Angular Ventures and Redseed."Source: [siliconangle.com](https://siliconangle.com/2025/04/22/cloud-security-startup-reco-raises-25m-funding/), read 2026-11-16. - **reco-24** (vendor): Reco's customers page links to Gartner Peer Insights and its newsroom lists a CRN 2025 Stellar Startup recognition; no Gartner or Forrester analyst report on Reco was found on 2026-11-16.Quote: "View Reco on Gartner Peer Insights"Source: [reco.ai](https://www.reco.ai/customers), read 2026-11-16. ## News mentioning Reco - [Reco raises $55 million as it widens from SaaS to agents](/news/reco-raises-55-million-agent-security/) ## Related pages - [All vendor profiles](/vendors/) - [Push Security profile](/vendors/push-security/) - [SentinelOne (Prompt Security) profile](/vendors/sentinelone/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # SentinelOne (Prompt Security) for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/sentinelone/ Updated: 2026-11-16 SentinelOne sells shadow AI detection as Prompt Security AI Usage Control. It uses a browser extension, a dedicated AI endpoint agent, an MCP gateway and API integrations, and says it recognizes more than 15,000 AI tools. Documented actions include inline redaction and in-the-moment coaching, with prompt blocking reported by Techzine, and an audit trail of prompts by user and department. The vendor says it deploys in minutes. A self-hosted on-premises edition exists. None found on 16 Nov 2026 for OAuth app discovery, identity provider sources, a per-tool risk score, a measured time to a first inventory or a Prompt Security price. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Large named AI tool library (15,000+) with browser and endpoint sensors, inline redaction and coaching (sentinelone-08, sentinelone-13, sentinelone-19, sentinelone-20). **Weakness:** None found on 2026-11-16 for OAuth app discovery, identity provider sources, per-tool risk ratings or a published price (sentinelone-12, sentinelone-16, sentinelone-24, sentinelone-26). ## Scores and evidence SentinelOne (Prompt Security) is 13th in the overall ranking at 48.0 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **SentinelOne (Prompt Security): score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | Vendor claims cover AI tools, named code assistants, agents, MCP servers, plugins, browsers and embedded copilots in general terms. None found for OAuth-granted AI apps or named embedded SaaS features. | [sentinelone-07](#fact-sentinelone-07), [sentinelone-09](#fact-sentinelone-09), [sentinelone-10](#fact-sentinelone-10), [sentinelone-11](#fact-sentinelone-11), [sentinelone-08](#fact-sentinelone-08), [sentinelone-12](#fact-sentinelone-12) | | Where it looks | 4 of 10 | 20 | Browser extension, endpoint agent, MCP gateway and API integrations are claimed. None found for identity provider, SaaS logs, email or billing. | [sentinelone-13](#fact-sentinelone-13), [sentinelone-14](#fact-sentinelone-14), [sentinelone-15](#fact-sentinelone-15), [sentinelone-16](#fact-sentinelone-16) | | Time to a first inventory | 4 of 10 | 15 | Vendor says it deploys in minutes with no infrastructure changes. None found for a measured time to inventory or an OS and browser list. | [sentinelone-17](#fact-sentinelone-17), [sentinelone-55](#fact-sentinelone-55), [sentinelone-18](#fact-sentinelone-18) | | What it can do about a finding | 6 of 10 | 15 | Inline redaction and in-the-moment coaching, with prompt blocking reported by Techzine. None found for OAuth revocation, approval or tickets. | [sentinelone-19](#fact-sentinelone-19), [sentinelone-20](#fact-sentinelone-20), [sentinelone-21](#fact-sentinelone-21), [sentinelone-24](#fact-sentinelone-24) | | Risk context and reporting | 4 of 10 | 10 | Audit trail of prompts and usage by user and department. None found for a per-tool risk rating or an exportable inventory. | [sentinelone-22](#fact-sentinelone-22), [sentinelone-23](#fact-sentinelone-23), [sentinelone-08](#fact-sentinelone-08), [sentinelone-24](#fact-sentinelone-24) | | Transparency | 3 of 10 | 10 | Published research and open-source code observed on GitHub. None found for a product price or public documentation. | [sentinelone-27](#fact-sentinelone-27), [sentinelone-28](#fact-sentinelone-28), [sentinelone-25](#fact-sentinelone-25), [sentinelone-26](#fact-sentinelone-26) | | Proof | 6 of 10 | 5 | Funding and acquisition reported by CRN and Techzine, named CISO quotes and a linked market report. | [sentinelone-03](#fact-sentinelone-03), [sentinelone-04](#fact-sentinelone-04), [sentinelone-05](#fact-sentinelone-05), [sentinelone-29](#fact-sentinelone-29), [sentinelone-30](#fact-sentinelone-30), [sentinelone-31](#fact-sentinelone-31) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 13th | 48.0 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 13th | 48.5 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 13th | 47.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | joint 13th | 50.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 14th | 48.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 13th | 51.5 | ## Product, deployment, price and owner - Product assessed: Prompt Security AI Usage Control (SentinelOne) - Deployment: Singularity Platform console; browser extension, dedicated AI endpoint agent, MCP Gateway and API integrations; self-hosted Prompt Security On-Premises; AgentCore gateway integration - Time to deploy: SentinelOne says it deploys in minutes; no measured figure or OS list found on 2026-11-16 - Price: none found on 2026-11-16 for Prompt Security; the pricing page lists Singularity tiers and says Enterprise is Contact Sales for Pricing - Owner: SentinelOne, Inc. (NYSE: S), Mountain View, California - Website: sentinelone.com ## Facts and sources - **sentinelone-01** (vendor): SentinelOne says Prompt Security is built into its Singularity Platform and managed from the same console as its endpoint, cloud, identity and data security.Quote: "Prompt Security is built into the SentinelOne Singularity Platform."Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-02** (vendor): On 2025-08-05 SentinelOne announced a definitive agreement to acquire Prompt Security, so the AI security line sits inside SentinelOne (NYSE: S).Quote: "signed a definitive agreement to acquire Prompt Security"Source: [sentinelone.com](https://www.sentinelone.com/press/sentinelone-to-acquire-prompt-security-to-advance-genai-security/), read 2026-11-16. - **sentinelone-03** (third-party): CRN reports that Prompt Security was founded in 2023 and raised an $18 million Series A in November 2024 before SentinelOne agreed to acquire it.Quote: "Founded in 2023, Prompt Security offers protection against AI-related threats"Source: [crn.com](https://www.crn.com/news/security/2025/sentinelone-to-acquire-ai-protection-startup-prompt-security/), read 2026-11-16. - **sentinelone-04** (third-party): CRN reports the Series A was led by Jump Capital with Hetz Ventures, Ridge Ventures, Okta and F5 taking part.Quote: "In November 2024, Prompt Security announced an $18 million Series A funding round led by Jump Capital"Source: [crn.com](https://www.crn.com/news/security/2025/sentinelone-to-acquire-ai-protection-startup-prompt-security/), read 2026-11-16. - **sentinelone-05** (third-party): Techzine reports SentinelOne acquired Israeli company Prompt Security to prepare its Singularity platform for secure AI use in organizations.Quote: "SentinelOne has announced the acquisition of Israeli company Prompt Security."Source: [techzine.eu](https://www.techzine.eu/news/security/133547/sentinelone-strengthens-ai-protection-with-prompt-security-deal/), read 2026-11-16. - **sentinelone-06** (third-party): Wikipedia lists SentinelOne as a public company headquartered in Mountain View, California, founded in 2013, with Tomer Weingarten as chairman and CEO.Quote: "Headquarters Mountain View, California , U.S."Source: [en.wikipedia.org](https://en.wikipedia.org/wiki/SentinelOne), read 2026-11-16. - **sentinelone-07** (vendor): SentinelOne says Prompt Security inventories every AI tool and code assistant in use, including unsanctioned shadow AI.Quote: "Inventory every AI tool and code assistant in use, including unsanctioned shadow AI"Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-08** (vendor): SentinelOne says AI Usage Control detects usage across more than 15,000 AI tools.Quote: "Detect usage across 15,000+ AI tools, the largest library in the category"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-09** (vendor): SentinelOne says Prompt Security tracks AI code assistants in the environment and names GitHub Copilot, Cursor and Claude Code.Quote: "Track every AI code assistant in the environment, including GitHub Copilot, Cursor, and Claude Code"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-10** (vendor): SentinelOne says Prompt Security discovers every AI agent and MCP server in the environment, including the connectors, tools, skills and plugins an agent touches.Quote: "See every connector, tool, skill, and plugin an agent touches, and who's running it"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-agentic-ai-security/), read 2026-11-16. - **sentinelone-11** (vendor): SentinelOne says AI Usage Control discovers and governs AI use across employees, browsers and embedded copilots, unsanctioned tools included.Quote: "Discover and govern AI usage across every employee, browser, and embedded copilot, unsanctioned tools included."Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-12** (none-found): None found on 2026-11-16: looked on the Prompt Security product page and AI Usage Control page for discovery of AI apps granted OAuth access to Google Workspace or Microsoft 365 and for named SaaS-embedded AI features (Notion AI, Slack AI, Zoom AI); only the general "embedded copilot" wording was found.Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-13** (vendor): Prompt Security says AI governance needs local sensors, namely a browser extension and a dedicated AI endpoint agent, because network controls alone fall short.Quote: "you need local, purpose-built sensors: a browser extension and a dedicated AI endpoint agent"Source: [prompt.security](https://prompt.security/blog/the-key-layer-in-ai-security-browser-and-endpoint-sensors), read 2026-11-16. - **sentinelone-14** (vendor): SentinelOne says an MCP Gateway, browser extension, security plugins and API integrations work together so no agent surface goes undetected.Quote: "MCP Gateway, browser extension, security plugins, and API integrations work together so no agent surface goes undetected"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-agentic-ai-security/), read 2026-11-16. - **sentinelone-15** (vendor): In its acquisition release SentinelOne says Prompt Security's technology integrates across browsers, desktop applications and APIs.Quote: "integrating across browsers, desktop applications, and APIs"Source: [sentinelone.com](https://www.sentinelone.com/press/sentinelone-to-acquire-prompt-security-to-advance-genai-security/), read 2026-11-16. - **sentinelone-16** (none-found): None found on 2026-11-16: looked on the Prompt Security, AI Usage Control and Agentic AI Security pages for identity provider (Okta, Entra ID, Google) and SaaS audit log, email or billing record integrations as discovery sources; none named.Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-17** (vendor): SentinelOne says Prompt Security deploys in minutes with no complex integrations or infrastructure changes.Quote: "Prompt Security deploys in minutes. No complex integrations or infrastructure changes are required."Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-18** (none-found): None found on 2026-11-16: looked on the Prompt Security product pages for supported operating systems, browsers, a measured time to first AI inventory and a published deployment requirements list; only "minutes" was found.Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-19** (vendor): SentinelOne says AI Usage Control redacts PII, secrets and other sensitive data inline on the device before a prompt leaves it.Quote: "Redact PII, secrets, and other sensitive data at the edge, inline, before a prompt ever leaves the device"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-20** (vendor): SentinelOne says AI Usage Control coaches employees in the moment instead of blocking their work.Quote: "Coach employees in the moment instead of blocking their work"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-21** (third-party): Techzine reports that Prompt Security uses policy-based controls, for example blocking risky prompts before they are sent.Quote: "Policy-based controls enforce safe AI usage by, for example, blocking risky prompts before they are sent."Source: [techzine.eu](https://www.techzine.eu/news/security/133547/sentinelone-strengthens-ai-protection-with-prompt-security-deal/), read 2026-11-16. - **sentinelone-22** (vendor): SentinelOne says AI Usage Control keeps a full audit trail of every prompt, redaction and policy action.Quote: "Review every prompt, redaction, and policy action in one place, with a full audit trail across every AI interaction"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-23** (vendor): SentinelOne says AI Usage Control shows every tool by user and department in one view.Quote: "See every tool, every user, every department, in one view"Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-24** (none-found): None found on 2026-11-16: looked on the AI Usage Control page for a per-tool risk rating (training on inputs, vendor posture), an exportable inventory report, and OAuth grant revocation or ticket routing as actions; none described.Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-25** (vendor): SentinelOne's public pricing page lists tiers but says the Enterprise tier is "Contact Sales for Pricing" and does not list Prompt Security.Quote: "Contact Sales for Pricing"Source: [sentinelone.com](https://www.sentinelone.com/pricing/), read 2026-11-16. - **sentinelone-26** (none-found): None found on 2026-11-16: looked on the pricing page and Prompt Security pages for a published price of Prompt Security or AI Usage Control and for public technical documentation readable without sign-in; found marketing pages and a demo form only.Source: [sentinelone.com](https://www.sentinelone.com/pricing/), read 2026-11-16. - **sentinelone-27** (vendor): On 2026-08-20 Prompt Security published research on three agent cases (AGENTS.md goal hijack, marketplace skills dependency hijack, instruction file integrity) and says approval prompts should show the real destination and payload.Quote: "approval prompts that show the real destination and payload instead of a friendly summary"Source: [prompt.security](https://prompt.security/blog/when-ai-agents-become-the-supply-chain-hidden-control-planes-in-agentic-systems), read 2026-11-16. - **sentinelone-28** (observed): Prompt Security publishes ClawSec on GitHub under AGPL-3.0, a security skill suite for OpenClaw, Hermes, PicoClaw and NanoClaw agents with 1.1k stars.Quote: "ClawSec source code is licensed under GNU AGPL-3.0-or-later"Source: [github.com](https://github.com/prompt-security/clawsec), read 2026-11-16. - **sentinelone-29** (vendor): SentinelOne's page quotes Margarita Rivera, Global CISO of Carnival Corporation, on Prompt Security giving a first layer of visibility and protection.Quote: "Prompt Security has been incredible for our organization."Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-30** (vendor): SentinelOne's page quotes Richard Moore, CISO at 10x Banking, on Prompt Security.Quote: "Prompt Security has allowed us to be enablers for the business."Source: [sentinelone.com](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/), read 2026-11-16. - **sentinelone-31** (vendor): SentinelOne's Prompt Security page links a "2026 Latio AI Security Market Report" among its resources (analyst-style coverage cited by the vendor; contents not read).Quote: "2026 Latio AI Security Market Report"Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. - **sentinelone-46** (vendor): On 2026-03-23 SentinelOne said Prompt Security On-Premises is self-hosted, works in disconnected environments, and blocks prompt injections and data leaks as a firewall for internal and external AI tools.Quote: "Prompt Security blocks threats like prompt injections and data leaks"Source: [sentinelone.com](https://www.sentinelone.com/press/sentinelone-brings-ai-security-to-on-premise-regulated-sovereign-self-hosted-and-airgapped-environments/), read 2026-11-16. - **sentinelone-55** (vendor): SentinelOne says setup is to connect your environment and integrate with existing infrastructure in minutes, with protection starting immediately.Quote: "Integrate with your existing infrastructure in minutes. No disruption. Protection starts immediately."Source: [sentinelone.com](https://www.sentinelone.com/platform/securing-ai-prompt/), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [Reco profile](/vendors/reco/) - [WitnessAI profile](/vendors/witnessai/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # WitnessAI for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/witnessai/ Updated: 2026-11-16 WitnessAI discovers shadow AI at the network level, agentless, with no endpoint client or browser extension. Observe draws on a catalog of more than 4,000 AI apps, including native desktop apps and coding agents, and the vendor lists agentic plugins, AI extensions and MCP servers. Control adds policy, real-time redaction and tokenizing, and denial of unapproved tools or MCP servers. An F5 BIG-IP SSL Orchestrator integration is documented. It reports who uses which tool and cost. None found on 16 Nov 2026 for OAuth app discovery, identity provider or SaaS API sources, a published time to a first inventory, public documentation or a price: the pricing page returned 404. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Agentless network-level discovery from a catalog of more than 4,000 AI apps, including native desktop apps and coding agents, with real-time redaction and tokenizing (witnessai-05, witnessai-06, witnessai-12). **Weakness:** None found on 2026-11-16 for OAuth app discovery, identity provider or SaaS API sources, a published time to first inventory, or a public price (witnessai-10, witnessai-11, witnessai-20). ## Scores and evidence WitnessAI is 16th in the overall ranking at 42.0 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **WitnessAI: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 6 of 10 | 25 | Network discovery of AI apps, native apps, named coding tools, agentic plugins, AI extensions and MCP servers is claimed. None found for OAuth apps or embedded SaaS AI. | [witnessai-05](#fact-witnessai-05), [witnessai-06](#fact-witnessai-06), [witnessai-07](#fact-witnessai-07), [witnessai-08](#fact-witnessai-08), [witnessai-28](#fact-witnessai-28), [witnessai-10](#fact-witnessai-10) | | Where it looks | 3 of 10 | 20 | Network only, with a named SSL orchestrator integration. None found for IdP, SaaS API, email or billing sources. | [witnessai-04](#fact-witnessai-04), [witnessai-06](#fact-witnessai-06), [witnessai-09](#fact-witnessai-09), [witnessai-10](#fact-witnessai-10) | | Time to a first inventory | 3 of 10 | 15 | Agentless network deployment is claimed. None found for a time to first inventory or an OS and browser list. | [witnessai-05](#fact-witnessai-05), [witnessai-09](#fact-witnessai-09), [witnessai-11](#fact-witnessai-11) | | What it can do about a finding | 5 of 10 | 15 | Policy enforcement, real-time redaction and denial of unapproved tools or MCP servers. None found for coaching, OAuth revocation or tickets. | [witnessai-12](#fact-witnessai-12), [witnessai-13](#fact-witnessai-13), [witnessai-14](#fact-witnessai-14) | | Risk context and reporting | 4 of 10 | 10 | Who uses which tool and cost, plus an MCP catalog risk score. None found for a per-app data-use rating or an exportable report. | [witnessai-15](#fact-witnessai-15), [witnessai-16](#fact-witnessai-16), [witnessai-17](#fact-witnessai-17) | | Transparency | 2 of 10 | 10 | A published survey report. None found for public documentation or a price. | [witnessai-21](#fact-witnessai-21), [witnessai-19](#fact-witnessai-19), [witnessai-20](#fact-witnessai-20) | | Proof | 6 of 10 | 5 | SecurityWeek reports over $85 million raised; a named case study and a vendor-hosted IDC excerpt. | [witnessai-03](#fact-witnessai-03), [witnessai-22](#fact-witnessai-22), [witnessai-23](#fact-witnessai-23), [witnessai-24](#fact-witnessai-24) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 16th | 42.0 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | 16th | 44.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 16th | 41.5 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 16th | 45.0 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 16th | 42.5 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | 16th | 45.0 | ## Product, deployment, price and owner - Product assessed: WitnessAI Observe (AI application and agent discovery) with Control for policy - Deployment: Network-level, agentless discovery and enforcement (no endpoint client or browser extension), F5 BIG-IP SSL Orchestrator integration, and REST APIs for custom agents - Time to deploy: none found on 2026-11-16 - Price: none found on 2026-11-16 (pricing page returns 404; demo request only) - Owner: WitnessAI (standalone, venture-backed; no parent company stated on its pages) - Website: [witness.ai](https://witness.ai) ## Facts and sources - **witnessai-01** (vendor): WitnessAI describes itself as an AI security and governance platform for employees, models, applications and agents, sold as Observe, Control and Protect (with Attack for red teaming).Quote: "WitnessAI makes AI safe to use and deploy for every employee, model, application, and agent."Source: [witness.ai](https://witness.ai/), read 2026-11-16. - **witnessai-02** (vendor): WitnessAI lists Rick Caccia as co-founder and CEO on its own leadership page.Quote: "Chief Executive Officer & Co-founder"Source: [witness.ai](https://witness.ai/about-us/), read 2026-11-16. - **witnessai-03** (third-party): SecurityWeek reported on 2026-01-13 that WitnessAI raised $58 million, bringing total funding to over $85 million, with a standalone investor round led by Sound Ventures.Quote: "announced raising $58 million in strategic funding, which brings the total raised by the company to over $85 million."Source: [securityweek.com](https://www.securityweek.com/witnessai-raises-58-million-for-ai-security-platform/), read 2026-11-16. - **witnessai-04** (vendor): WitnessAI says it scans the network for third-party AI applications and agents to give visibility into AI adoption.Quote: "Scan your entire network for third-party AI applications and agents for complete visibility into AI adoption and risk exposure"Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-05** (vendor): WitnessAI says its catalog can detect more than 4,000 AI applications without an endpoint client.Quote: "AI applications can be detected by the WitnessAI catalog, no endpoint client required."Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-06** (vendor): WitnessAI says it sees AI activity at the network level without browser extensions or endpoint clients, including native apps such as Windows Copilot and Office 365.Quote: "See AI activity across your entire network without relying on browser extensions or endpoint clients."Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-07** (vendor): WitnessAI says it discovers AI developer tools such as GitHub Copilot and Cursor across the network and shows which coding agents are running.Quote: "Discover apps like GitHub Copilot, Cursor, and hundreds of other AI dev tools across your network."Source: [witness.ai](https://witness.ai/for-developers/), read 2026-11-16. - **witnessai-08** (vendor): WitnessAI says it shows which agentic plugins employees install and which MCP servers they reach.Quote: "See which agentic plugins employees install and which MCP servers they reach."Source: [witness.ai](https://witness.ai/for-employees/), read 2026-11-16. - **witnessai-09** (vendor): WitnessAI says it works over network connectivity and has a named integration with F5 BIG-IP SSL Orchestrator to see encrypted AI traffic.Quote: "WitnessAI uses network connectivity to observe, control, and protect employees' use of AI applications."Source: [witness.ai](https://witness.ai/resources/witnessai-announces-integration-with-f5-big-ip-to-deliver-visibility-and-control-over-encrypted-genai-activity/), read 2026-11-16. - **witnessai-10** (none-found): None found on 2026-11-16: named identity provider, SaaS API, email or billing record sources, OAuth grant discovery, and embedded SaaS AI feature discovery on the Observe, Product and For Employees pages.Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-11** (none-found): None found on 2026-11-16: a published time to first AI inventory, or a list of supported operating systems and browsers, on the Observe and Product pages.Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-12** (vendor): WitnessAI says it enforces policies, redacts sensitive data in real time and writes audit trails.Quote: "Intelligently route prompts and enforce policies, redact sensitive data in real time, and generate granular audit trails."Source: [witness.ai](https://witness.ai/control/), read 2026-11-16. - **witnessai-13** (vendor): WitnessAI says it identifies and tokenizes PII, credentials and secrets before they reach an AI model or agent.Quote: "identifying and tokenizing them before they ever reach an AI model or agent."Source: [witness.ai](https://witness.ai/protect/), read 2026-11-16. - **witnessai-14** (vendor): WitnessAI says a tool or MCP server not on the approved list is denied before the tool executes, whichever model provider the call is routed through.Quote: "Anything off the list is denied before the tool executes"Source: [witness.ai](https://witness.ai/blog/introducing-witnessai-agentic-control-one-control-plane-for-every-agent-tool-and-mcp-server/), read 2026-11-16. - **witnessai-15** (vendor): WitnessAI says it shows who is using which AI tool, what for and at what cost, and catalogs AI apps and agent activity for reporting.Quote: "Understand who, what, why, and cost for every AI interaction, and track trends across your entire workforce."Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-16** (vendor): On 2026-06-17 WitnessAI announced an MCP Catalog that scores known tools against OWASP and CVE risk classes so teams can assess risk before approval.Quote: "WitnessAI's new MCP Catalog scores known tools against OWASP and CVE risk classes, enabling security teams to assess risk before granting approval."Source: [prnewswire.com](https://www.prnewswire.com/news-releases/witnessai-introduces-agentic-control-to-secure-and-govern-ai-agents-and-mcp-servers-302802606.html), read 2026-11-16. - **witnessai-17** (none-found): None found on 2026-11-16: a per-application risk rating for web AI apps (data use, training on inputs, vendor posture) or an exportable inventory report, on the Observe and Product pages.Source: [witness.ai](https://witness.ai/observe/), read 2026-11-16. - **witnessai-18** (vendor): On 2026-09-15 WitnessAI announced AI FinOps capabilities that combine multi-provider metering, model routing, shadow AI discovery and runtime security.Quote: "multi-provider metering, intelligent model routing, shadow AI discovery, and runtime security"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/witnessai-introduces-ai-finops-capabilities-to-control-enterprise-ai-spend-and-drive-effective-roi-302878268.html), read 2026-11-16. - **witnessai-19** (none-found): None found on 2026-11-16: public documentation: docs.witness.ai loaded as a password-gated page ("Please enter your password for this site") with no readable documentation.Source: [docs.witness.ai](https://docs.witness.ai/), read 2026-11-16. - **witnessai-20** (none-found): None found on 2026-11-16: a published price: https://witness.ai/pricing/ returned a 404 page, and no price appeared on the Product, Observe, Control or Protect pages; the pages direct buyers to book a demo.Source: [witness.ai](https://witness.ai/pricing/), read 2026-11-16. - **witnessai-21** (vendor): On 2026-07-22 WitnessAI published a survey report, The Hidden Cost of Enterprise AI, finding that few enterprises formally inventory and approve all agents.Quote: "fewer than one in five (18%) report that all agents are formally inventoried and approved by their security team"Source: [prnewswire.com](https://www.prnewswire.com/news-releases/new-witnessai-report-reveals-the-financial-cost-and-risk-factors-of-enterprise-ai-adoption-302831518.html), read 2026-11-16. - **witnessai-22** (vendor): WitnessAI publishes a customer case study with InComm Payments, with its CISO quoted on compliance, data-loss prevention and privacy visibility.Quote: "Our compliance, data-loss prevention, and privacy teams now have total visibility and confidence in our AI security."Source: [witness.ai](https://witness.ai/resources/incomm-payments/), read 2026-11-16. - **witnessai-23** (vendor): WitnessAI says it was named a founding member of the ServiceNow Control Tower Alliance.Quote: "WitnessAI Named a Founding Member of ServiceNow Control Tower Alliance"Source: [witness.ai](https://witness.ai/for-developers/), read 2026-11-16. - **witnessai-24** (vendor): WitnessAI hosts an excerpt of IDC Innovators: Security for Agentic AI, 2025, an analyst document covering emerging vendors in this area.Quote: "This IDC Innovators presentation highlights emerging vendors addressing the security challenges of agentic AI"Source: [witness.ai](https://witness.ai/resources/idc-innovators-security-for-agentic-ai-2025/), read 2026-11-16. - **witnessai-25** (vendor): WitnessAI says it captures prompts, tool calls and shell executions across supported Cursor, Claude Code and Codex workflows.Quote: "Capture prompts, tool calls, and shell executions across supported Cursor, Claude Code, and Codex workflows."Source: [witness.ai](https://witness.ai/for-agents/), read 2026-11-16. - **witnessai-28** (vendor): WitnessAI says it discovers agentic activity in VS Code with AI extensions and ChatGPT with enabled plugins; this is about AI extensions and plugins, not a described AI-browser product.Quote: "VSCode with AI extensions, ChatGPT with enabled plugins, and local agents running on developer machines"Source: [witness.ai](https://witness.ai/blog/introducing-witnessai-agentic-security-extending-the-confidence-layer-to-ai-agents/), read 2026-11-16. - **witnessai-33** (vendor): WitnessAI says custom agents get runtime protection through lightweight REST APIs added to the orchestration layer.Quote: "Add runtime protection to agent orchestration through lightweight REST APIs."Source: [witness.ai](https://witness.ai/for-agents/), read 2026-11-16. - **witnessai-39** (none-found): None found on 2026-11-16: a published time to deploy, system requirements, or supported operating systems on the Product, For Agents and Control pages.Source: [witness.ai](https://witness.ai/product/), read 2026-11-16. ## Related pages - [All vendor profiles](/vendors/) - [SentinelOne (Prompt Security) profile](/vendors/sentinelone/) - [Zscaler profile](/vendors/zscaler/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Zscaler for shadow AI detection URL: https://shadowaidetectiontools.com/vendors/zscaler/ Updated: 2026-11-16 Zscaler reads shadow AI from inline traffic through its Zero Trust Exchange and from an endpoint layer. Its AI Asset Management documentation keeps separate inventories for GenAI web apps, embedded AI in SaaS and endpoint AI apps, plus an MCP server inventory that can be exported to Excel. Each app carries a risk index covering hosting, posture and retention, with usage by user and department. Controls include blocking by role, data filtering, browser isolation and coaching. Documentation is readable without sign-in. None found on 16 Nov 2026 for a published price, a time to a first inventory, an operating system list, or discovery of OAuth grants and identity provider sources. ## Strength and weakness for Shadow AI Detection Tools **Strength:** Separate documented inventories for GenAI web apps, embedded SaaS AI and endpoint AI apps, with a per-app risk index, from inline network traffic plus an endpoint layer (zscaler-08, zscaler-09, zscaler-10, zscaler-11). **Weakness:** None found on 2026-11-16: published price, time to first inventory, OS list, or OAuth-grant and identity provider discovery (zscaler-24, zscaler-25, zscaler-26, zscaler-27). ## Scores and evidence Zscaler is 6th in the overall ranking at 56.5 of 100. Each score below lists the fact ids behind it; the facts follow further down with their sources and read dates. **Zscaler: score per criterion** | Criterion | Score | Weight | Why | Facts | |---|---|---|---|---| | What it discovers | 7 of 10 | 25 | Vendor documents shadow GenAI apps, embedded SaaS AI, endpoint AI apps, agents, MCP servers, extensions and dev tools; the only second source is the vendor's own annual report. None found for OAuth-granted apps. | [zscaler-04](#fact-zscaler-04), [zscaler-10](#fact-zscaler-10), [zscaler-11](#fact-zscaler-11), [zscaler-15](#fact-zscaler-15), [zscaler-03](#fact-zscaler-03), [zscaler-26](#fact-zscaler-26) | | Where it looks | 6 of 10 | 20 | Inline proxy, endpoint and cloud environments are claimed. None found for IdP, email or billing sources. | [zscaler-07](#fact-zscaler-07), [zscaler-11](#fact-zscaler-11), [zscaler-15](#fact-zscaler-15), [zscaler-13](#fact-zscaler-13), [zscaler-26](#fact-zscaler-26) | | Time to a first inventory | 1 of 10 | 15 | Only an inline position is described. None found for a time to first inventory, setup type or OS list. | [zscaler-07](#fact-zscaler-07), [zscaler-25](#fact-zscaler-25) | | What it can do about a finding | 6 of 10 | 15 | Block by role, data filtering, browser isolation, endpoint restriction and coaching, vendor documented. None found for OAuth revocation, approval flows or tickets. | [zscaler-05](#fact-zscaler-05), [zscaler-12](#fact-zscaler-12), [zscaler-16](#fact-zscaler-16), [zscaler-19](#fact-zscaler-19), [zscaler-20](#fact-zscaler-20), [zscaler-03](#fact-zscaler-03), [zscaler-27](#fact-zscaler-27) | | Risk context and reporting | 7 of 10 | 10 | Per-app risk index with hosting, posture and retention, usage by user and department, and Excel export for MCP inventory; vendor-only. | [zscaler-09](#fact-zscaler-09), [zscaler-08](#fact-zscaler-08), [zscaler-21](#fact-zscaler-21), [zscaler-14](#fact-zscaler-14) | | Transparency | 6 of 10 | 10 | Documentation observed readable without sign-in and published research. None found for a price. | [zscaler-23](#fact-zscaler-23), [zscaler-22](#fact-zscaler-22), [zscaler-24](#fact-zscaler-24) | | Proof | 7 of 10 | 5 | The vendor's annual report states about 11,000 customers, with named customer quotes and claimed analyst leader rankings; Wikipedia confirms only the company. None found for an independent analyst report. | [zscaler-02](#fact-zscaler-02), [zscaler-28](#fact-zscaler-28), [zscaler-29](#fact-zscaler-29), [zscaler-31](#fact-zscaler-31) | ## Where it ranks | View | Rank | Total | |---|---|---| | [Overall ranking](/ranking/) | 6th | 56.5 | | [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) | joint 4th | 63.0 | | [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) | 11th | 49.0 | | [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) | 5th | 59.5 | | [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) | 6th | 59.0 | | [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) | joint 6th | 58.5 | ## Product, deployment, price and owner - Product assessed: Zscaler AI Security: AI Asset Management (Public AI inventory for GenAI, embedded AI, endpoint AI, MCP servers), Endpoint AI Security, and Zscaler Internet Access GenAI controls (Zero Trust Exchange) - Deployment: Inline cloud proxy (Zero Trust Exchange) for web and GenAI traffic, plus an endpoint layer for Endpoint AI Security; setup steps not read - Time to deploy: none found on 2026-11-16 - Price: none found on 2026-11-16 - Owner: Zscaler, Inc. (Nasdaq: ZS), a standalone public company - Website: zscaler.com ## Facts and sources - **zscaler-01** (third-party): Zscaler, Inc. was incorporated in 2007 as SafeChannel and renamed Zscaler in 2008; its 10-K filed 2026-09-03 lists San Jose, California as its headquarters, and no parent company is named.Quote: "We were incorporated in the state of Delaware in September 2007 as SafeChannel, Inc., and in August 2008, we changed our name to Zscaler, Inc."Source: [sec.gov](https://www.sec.gov/Archives/edgar/data/1713683/000171368326000157/zs-20260731.htm), read 2026-11-16. - **zscaler-02** (third-party): In its annual report filed 2026-09-03, Zscaler says it has about 11,000 customers, including over 40% of the Forbes Global 2000.Quote: "We have approximately 11,000 customers globally, including many of the largest enterprises and government agencies across every major industry."Source: [sec.gov](https://www.sec.gov/Archives/edgar/data/1713683/000171368326000157/zs-20260731.htm), read 2026-11-16. - **zscaler-03** (third-party): On 2026-09-03, Zscaler's 10-K listed discovering and remediating unsanctioned AI tools and shadow AI deployments as a primary use case of its Security for AI solution.Quote: "discovering and remediating unsanctioned AI tools and shadow AI deployments before they introduce unmanaged risk"Source: [sec.gov](https://www.sec.gov/Archives/edgar/data/1713683/000171368326000157/zs-20260731.htm), read 2026-11-16. - **zscaler-04** (vendor): Zscaler's AI Security page says AI Asset Management auto-discovers models, agents, MCP servers and shadow AI across 2,900+ apps in one view.The exact quote is on the source page.Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16. - **zscaler-05** (vendor): Zscaler says its Secure Access to AI capability blocks AI access by role in real time and filters sensitive data sent to LLMs.Quote: "Control who uses AI and how: block access by role in real time, filter sensitive data from LLMs, and govern agent-to-agent traffic."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16. - **zscaler-06** (vendor): Zscaler says it can find and stop AI threats on employee devices in browsers, extensions and plugins.Quote: "Find and stop AI threats on employee devices in browsers, extensions, and plugins that traditional EDRs were never built to see."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16. - **zscaler-07** (vendor): Zscaler says it sits inline across the AI workflow, which means network-path (proxy) visibility, but this page gives no setup time.The exact quote is on the source page.Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16. - **zscaler-08** (vendor): Zscaler documents a Public AI Inventory that covers sanctioned and unsanctioned (shadow) GenAI apps, showing who uses which app and how often.Quote: "provides visibility into sanctioned and unsanctioned (shadow) AI usage who is accessing which apps and how often"Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai), read 2026-11-16. - **zscaler-09** (vendor): Zscaler documents a per-app risk index and risk attributes (hosting, security posture, data retention) for each GenAI app in the inventory.Quote: "Surfaces app risk signals, e.g., risk index to prioritize review and policy actions."Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai), read 2026-11-16. - **zscaler-10** (vendor): Zscaler documents an Embedded AI inventory that automatically discovers AI features built into SaaS apps employees use.Quote: "automatically discover and identify Embedded AI applications accessed across your organization"Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai), read 2026-11-16. - **zscaler-11** (vendor): Zscaler documents an Endpoint AI inventory of AI apps installed or used on endpoints, with instance counts and categories.Quote: "giving you a structured, continuously updated view of the AI applications installed or used on endpoints."Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-endpoint-ai), read 2026-11-16. - **zscaler-12** (vendor): Zscaler documents an overview showing per-app policy actions (allowed, blocked or isolated), usage by department and top users, and drill-down for incident response.Quote: "View the top 5 Gen AI apps used in the organization with policy actions (**Allowed**, **Blocked**, or **Isolated**)"Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-overview), read 2026-11-16. - **zscaler-13** (vendor): Zscaler documents an MCP server inventory with risk ratings and Excel export, but the sources shown are cloud environments such as AWS, Azure and GCP rather than developer machines.Quote: "Returns data about the environment where the MCP server was discovered, eg. AWS, Azure, or GCP."Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-ai-assets-mcp-servers), read 2026-11-16. - **zscaler-14** (vendor): Zscaler documents that the MCP server inventory can be exported as an Excel file.Quote: "Click **Export**to export the data and download the report as an Excel file."Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-ai-assets-mcp-servers), read 2026-11-16. - **zscaler-15** (vendor): On 2026-10-05, Zscaler said its Endpoint AI Security identifies AI assistants, agents, browsers, development tools, models, packages, extensions and connected services on managed devices.Quote: "identifies AI assistants, agents, browsers, development tools, models, packages, extensions, and connected services used across managed devices"Source: [zscaler.com](https://www.zscaler.com/blogs/product-insights/bringing-anthropic-s-claude-mythos-5-1-zscaler-endpoint-ai-security), read 2026-11-16. - **zscaler-16** (vendor): On 2026-10-05, Zscaler said endpoint policies can restrict unauthorized agents and tools, stop dangerous commands and protect sensitive files.Quote: "Policies can restrict unauthorized agents and tools, stop dangerous commands or instructions, and protect sensitive files and data."Source: [zscaler.com](https://www.zscaler.com/blogs/product-insights/bringing-anthropic-s-claude-mythos-5-1-zscaler-endpoint-ai-security), read 2026-11-16. - **zscaler-17** (vendor): On 2026-10-16, Zscaler's AI Security release notes added MCP server posture policies such as critical risk level and exposed code execution tools.Quote: "MCP server exposes code execution tools"Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/release-upgrade-summary-2026), read 2026-11-16. - **zscaler-18** (vendor): Zscaler says it can enforce which AI tools users can and cannot access.Quote: "Define and enforce which AI tools your users can and cannot access."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/securing-generative-ai), read 2026-11-16. - **zscaler-19** (vendor): Zscaler says it can disable cut, paste and download in an isolated browser to stop data loss to GenAI apps.Quote: "Stop data loss by disabling cut, paste, and download actions via a secure isolated browser."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/securing-generative-ai), read 2026-11-16. - **zscaler-20** (vendor): Zscaler says it pairs with Workflow Automation to coach users on AI usage.Quote: "Pair with Zscaler Workflow Automation to coach users on AI usage best practices."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/securing-generative-ai), read 2026-11-16. - **zscaler-21** (vendor): Zscaler says it shows which AI apps are used by user, department and organization, and the prompts users send.Quote: "See which AI apps are in use across your users, departments, and organization with interactive dashboards."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/securing-generative-ai), read 2026-11-16. - **zscaler-22** (vendor): Zscaler ThreatLabz's 2026 AI Security report blog warns that AI embedded in SaaS creates data pathways that teams often do not monitor.Quote: "Embedded AI may feel like a simple feature enhancement, but it often introduces new data pathways."Source: [zscaler.com](https://www.zscaler.com/blogs/security-research/ai-now-default-enterprise-accelerator-takeaways-threatlabz-2026-ai-security), read 2026-11-16. - **zscaler-23** (observed): Zscaler's AI Security product documentation is readable without sign-in at help.zscaler.com.Quote: "Public AI Overview page in AI Security summarizes usage across Gen AI apps and Endpoint AI apps"Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-overview), read 2026-11-16. - **zscaler-24** (none-found): None found on 2026-11-16: a published price for Zscaler AI Security, AI Asset Management or Endpoint AI Security; the /products-and-solutions/pricing page returned page not found.Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/pricing), read 2026-11-16. - **zscaler-25** (none-found): None found on 2026-11-16: a published time to first AI inventory, agentless versus agent setup, or operating system and browser list for Endpoint AI Security, on the product, docs and blog pages read.Source: [zscaler.com](https://www.zscaler.com/blogs/product-insights/bringing-anthropic-s-claude-mythos-5-1-zscaler-endpoint-ai-security), read 2026-11-16. - **zscaler-26** (none-found): None found on 2026-11-16: identity provider (Okta, Entra ID, Google) or OAuth-grant discovery of AI apps, or email or billing records as a discovery source, in the Zscaler AI inventory docs read.Source: [help.zscaler.com](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-overview), read 2026-11-16. - **zscaler-27** (none-found): None found on 2026-11-16: revoking OAuth grants or routing a discovered AI app to a ticket, on the Zscaler GenAI security page.Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/securing-generative-ai), read 2026-11-16. - **zscaler-28** (vendor): Zscaler's AI Security page quotes customer CISOs and IT directors, naming Repsol, Inter and BorgWarner.Quote: "Thanks to Zscaler, we can fully embrace the potential of GenAI."Source: [zscaler.com](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16. - **zscaler-29** (vendor): Zscaler says on its site that Gartner, Forrester and others rank it a leader; no specific shadow AI analyst report was read.Quote: "See why Gartner, Forrester, and others rank Zscaler a leader."Source: [zscaler.com](https://www.zscaler.com/company), read 2026-11-16. - **zscaler-30** (vendor): Zscaler's leadership page lists Jay Chaudhry as CEO, Chairman and Founder.Quote: "CEO, Chairman and Founder"Source: [zscaler.com](https://www.zscaler.com/company/leadership), read 2026-11-16. - **zscaler-31** (third-party): Wikipedia describes Zscaler as an American cybersecurity company, formerly SafeChannel, Inc., with Jay Chaudhry as a founder.Quote: "American cybersecurity company"Source: [en.wikipedia.org](https://en.wikipedia.org/wiki/Zscaler), read 2026-11-16. ## News mentioning Zscaler - [Zscaler adds MCP server policies to AI Asset Management](/news/zscaler-ai-asset-management-adds-mcp-server-policies/) - [Zscaler describes Endpoint AI Security for shadow AI](/news/zscaler-endpoint-ai-security-discovers-ai-on-managed-devices/) - [Zscaler adds AI Security policies for cloud agent platforms](/news/zscaler-adds-ai-security-policies-for-bedrock-foundry-copilot-studio/) ## Related pages - [All vendor profiles](/vendors/) - [WitnessAI profile](/vendors/witnessai/) - [Aim Security profile](/vendors/aim-security/) - [The ranking](/ranking/) - [How scores are set](/method/) - [Best tool by situation](/best-for/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Best shadow AI detection tool by situation URL: https://shadowaidetectiontools.com/best-for/ Updated: 2026-11-16 Five buyer situations reweight the same seven scores for shadow AI detection: a regulated inventory, a 30-day board deadline, a developer-heavy organization, SaaS-embedded AI, and coaching instead of blocking. Weights were fixed on 16 Nov 2026, and Harmonic Security leads the main ranking. ## 5 situations - [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/): Harmonic Security leads at 76.0; Bay is joint 13th. - [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/): Harmonic Security leads at 73.5; Bay is joint 6th. - [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/): Harmonic Security leads at 72.5; Bay is joint 10th. - [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/): Harmonic Security leads at 73.5; Bay is 12th. - [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/): Harmonic Security leads at 75.5; Bay is 11th. ## Who should pick what With no particular situation, the published ranking applies and Harmonic Security comes first. For a specific situation the weights change, and so can the order: - **[A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/)**: pick Harmonic Security (76.0). Bay is joint 13th at 48.5. An auditor asks for a complete inventory with risk ratings and an evidence trail, so breadth of discovery and the risk report weigh most. - **[A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/)**: pick Harmonic Security (73.5). Bay is joint 6th at 55.0. A hard deadline rewards the widest discovery that is running fastest; acting on findings can follow. - **[An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/)**: pick Harmonic Security (72.5). Bay is joint 10th at 52.0. Developer tools live on the endpoint and in the IDE, out of sight of SaaS logs, so what is discovered and where the tool looks matter most. - **[A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/)**: pick Harmonic Security (73.5). Bay is 12th at 50.0. Embedded AI and OAuth grants show up in identity and SaaS records, so reading those vantage points weighs most. - **[A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/)**: pick Harmonic Security (75.5). Bay is 11th at 53.0. Guiding people in the moment is the job, so what the tool can do when it finds an unapproved app weighs most. ## Related pages - [The overall ranking](/ranking/) - [Weights per situation](/method/) - [Vendor profiles](/vendors/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Best shadow AI detection tool for a regulated company URL: https://shadowaidetectiontools.com/best-for/regulated-inventory/ Updated: 2026-11-16 Harmonic Security leads when a regulated company must show auditors every AI tool touching customer data. The view weights what a tool discovers at 25, risk context and reporting at 25, and where it looks at 20, so breadth and an evidence trail count most. Weights fixed 16 Nov 2026. ## The short answer Harmonic Security leads the view "A regulated company that must show auditors every AI tool touching customer data" at 76.0 of 100. Bay is joint 13th at 48.5. Harmonic Security is ahead mainly on risk context and reporting (9 against 4, worth 12.5 points at weight 25). ## Why these weights An auditor asks for a complete inventory with risk ratings and an evidence trail, so breadth of discovery and the risk report weigh most. **Weights used in this view** | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---| | 25 | 20 | 5 | 15 | 25 | 5 | 5 | **A regulated company that must show auditors every AI tool touching customer data: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 25) | Where it looks (weight 20) | Time to a first inventory (weight 5) | What it can do about a finding (weight 15) | Risk context and reporting (weight 25) | Transparency (weight 5) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 76.0 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | 2nd | [Netskope](/vendors/netskope/) | 67.5 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 3rd | [Microsoft](/vendors/microsoft/) | 66.5 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | joint 4th | [Nudge Security](/vendors/nudge-security/) | 63.0 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | joint 4th | [Zscaler](/vendors/zscaler/) | 63.0 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | 6th | [Cyberhaven](/vendors/cyberhaven/) | 59.5 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | 7th | [Aurascape](/vendors/aurascape/) | 58.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | joint 8th | [Grip Security](/vendors/grip-security/) | 55.0 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | joint 8th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 55.0 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 10th | [Aim Security](/vendors/aim-security/) | 51.0 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 11th | [Reco](/vendors/reco/) | 50.5 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 12th | [Push Security](/vendors/push-security/) | 50.0 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | joint 13th | [Bay](/vendors/bay/) | 48.5 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | joint 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 48.5 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 15th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 47.5 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 44.0 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## What would change this ranking **Top pair.** Harmonic Security falling from 7 to 3 on what it discovers (weight 25, so each point moves a total by 2.5) would put Netskope above Harmonic Security; the gap is 8.5 points. **Bay's next place.** Bay rising from 6 to 7 on what it discovers (weight 25, so each point moves a total by 2.5) would put Bay above Push Security; the gap is 1.5 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which shadow AI detection tool is best for a regulated company? Harmonic Security leads this view. An auditor asks for a complete inventory with a risk assessment per tool and an evidence trail, so the weights put 25 on discovery, 20 on vantage points and 25 on risk context and reporting. Netskope and Microsoft follow the leader in this view. Netskope documents risk scores that include training on customer data and an inventory with usage, and Microsoft documents a catalog of more than 31,000 apps scored on 90-plus risk factors. Bay's position is joint 11th in the main ranking. ## What should a regulated buyer check before choosing? Ask each vendor for the exportable inventory or audit report, because export is a separate fact in each file. Harmonic's docs describe CSV export, Zscaler documents an Excel export for its MCP inventory, and for several tools none was found on 16 Nov 2026. Check retention of the detection records and whether the risk score is the vendor's catalog data or computed from your own traffic. The auditor guide on this site lists what evidence auditors ask for. ## Related pages - [All situations](/best-for/) - [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) - [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) - [Overall ranking](/ranking/) - [Method and weights](/method/) - [Vendor profiles](/vendors/) --- # Best shadow AI detection tool for a 30-day board inventory URL: https://shadowaidetectiontools.com/best-for/board-30-days/ Updated: 2026-11-16 Harmonic Security leads when a CISO must give the board an inventory of AI in use within 30 days. The view weights discovery at 30, time to a first inventory at 30 and where it looks at 20, since a hard deadline rewards wide discovery that runs fast. Weights fixed 16 Nov 2026. ## The short answer Harmonic Security leads the view "A CISO who must hand the board an inventory of AI in use within 30 days" at 73.5 of 100. Bay is joint 6th at 55.0. Harmonic Security is ahead mainly on where it looks (7 against 4, worth 6.0 points at weight 20). ## Why these weights A hard deadline rewards the widest discovery that is running fastest; acting on findings can follow. **Weights used in this view** | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---| | 30 | 20 | 30 | 5 | 5 | 5 | 5 | **A CISO who must hand the board an inventory of AI in use within 30 days: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 30) | Where it looks (weight 20) | Time to a first inventory (weight 30) | What it can do about a finding (weight 5) | Risk context and reporting (weight 5) | Transparency (weight 5) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 73.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | 2nd | [Microsoft](/vendors/microsoft/) | 67.5 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | 3rd | [Nudge Security](/vendors/nudge-security/) | 65.5 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 4th | [Netskope](/vendors/netskope/) | 63.5 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 5th | [Grip Security](/vendors/grip-security/) | 61.0 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | joint 6th | [Bay](/vendors/bay/) | 55.0 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | joint 6th | [Reco](/vendors/reco/) | 55.0 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 8th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 54.0 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 9th | [Push Security](/vendors/push-security/) | 52.5 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | 10th | [Cyberhaven](/vendors/cyberhaven/) | 51.5 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | 11th | [Zscaler](/vendors/zscaler/) | 49.0 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | 12th | [Aurascape](/vendors/aurascape/) | 48.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 47.5 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 14th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 46.5 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 15th | [Aim Security](/vendors/aim-security/) | 45.5 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 41.5 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## What would change this ranking **Top pair.** Harmonic Security falling from 7 to 4 on what it discovers (weight 30, so each point moves a total by 3.0) would put Microsoft above Harmonic Security; the gap is 6.0 points. **Bay's next place.** Bay rising from 6 to 9 on what it discovers (weight 30, so each point moves a total by 3.0) would put Bay above Grip Security; the gap is 6.0 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which tool is best when a board wants an inventory in 30 days? Harmonic Security leads this view. Speed carries 30 of the weights, and Harmonic says first insight arrives in about 30 minutes through MDM deployment, while the fastest published setups are API-based for identity and email tools. Microsoft and Nudge Security follow. Microsoft's local agent discovery starts automatically after Defender for Endpoint onboarding, and Nudge Security says most organizations deploy in under an hour from one read-only email connection. Grip Security says 10 minutes of setup and about 5 days for historic usage. ## What does a 30-day plan leave out? Acting on findings weighs only 5 here, so blocking and coaching can follow the first inventory. A fast API tool sees SaaS and OAuth use but not local apps, and an endpoint tool sees local apps but not every SaaS grant. Plan a second source in week two so the board number does not rest on one vantage point. The board metrics guide lists which numbers belong in the report. ## Related pages - [All situations](/best-for/) - [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) - [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) - [Overall ranking](/ranking/) - [Method and weights](/method/) - [Vendor profiles](/vendors/) --- # Best shadow AI detection tool for developer-heavy teams URL: https://shadowaidetectiontools.com/best-for/developer-heavy/ Updated: 2026-11-16 Harmonic Security leads for an engineering organization where developers add coding agents, IDE plugins and MCP servers on their own. The view weights discovery at 35 and where the tool looks at 25, because developer tools live on the endpoint and in the IDE, out of sight of SaaS logs. ## The short answer Harmonic Security leads the view "An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own" at 72.5 of 100. Bay is joint 10th at 52.0. Harmonic Security is ahead mainly on where it looks (7 against 4, worth 7.5 points at weight 25). ## Why these weights Developer tools live on the endpoint and in the IDE, out of sight of SaaS logs, so what is discovered and where the tool looks matter most. **Weights used in this view** | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---| | 35 | 25 | 10 | 15 | 5 | 5 | 5 | **An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 35) | Where it looks (weight 25) | Time to a first inventory (weight 10) | What it can do about a finding (weight 15) | Risk context and reporting (weight 5) | Transparency (weight 5) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 72.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | joint 2nd | [Microsoft](/vendors/microsoft/) | 69.0 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | joint 2nd | [Netskope](/vendors/netskope/) | 69.0 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 4th | [Nudge Security](/vendors/nudge-security/) | 64.0 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 5th | [Zscaler](/vendors/zscaler/) | 59.5 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | 6th | [Grip Security](/vendors/grip-security/) | 59.0 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | 7th | [Cyberhaven](/vendors/cyberhaven/) | 56.0 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | 8th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 55.5 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 9th | [Aurascape](/vendors/aurascape/) | 54.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | joint 10th | [Bay](/vendors/bay/) | 52.0 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | joint 10th | [Push Security](/vendors/push-security/) | 52.0 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | joint 10th | [Reco](/vendors/reco/) | 52.0 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | joint 13th | [Aim Security](/vendors/aim-security/) | 50.5 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | joint 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 50.5 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 15th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 47.0 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 45.0 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## What would change this ranking **Top pair.** Microsoft rising from 8 to 10 on what it discovers (weight 35, so each point moves a total by 3.5) would put Microsoft above Harmonic Security; the gap is 3.5 points. **Bay's next place.** Bay rising from 6 to 7 on what it discovers (weight 35, so each point moves a total by 3.5) would put Bay above Aurascape; the gap is 2.5 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which shadow AI detection tool is best for developers? Harmonic Security leads this view. Harmonic's endpoint page names desktop apps, coding assistants, command-line tools, local models and MCP servers, and its agent runs on Windows, macOS and Linux. Microsoft and Netskope share the next position. Microsoft's support matrix names Claude Code, Claude Desktop, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf and Kiro for local discovery on Windows and macOS. Netskope reads endpoint AI through its client and the gateway. ## Which tools miss developer tools? Identity and email tools such as Grip Security, Reco and Nudge Security document no endpoint vantage point, and Grip's file shows none found for MCP servers and coding agents on endpoints. Network tools see traffic to AI services but not local models or configuration files. For developer fleets, confirm the operating systems covered and whether MCP servers are listed per machine. ## Related pages - [All situations](/best-for/) - [A CISO who must hand the board an inventory of AI in use within 30 days](/best-for/board-30-days/) - [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) - [Overall ranking](/ranking/) - [Method and weights](/method/) - [Vendor profiles](/vendors/) --- # Best shadow AI detection tool for AI embedded in SaaS URL: https://shadowaidetectiontools.com/best-for/saas-embedded/ Updated: 2026-11-16 Harmonic Security leads for a SaaS-heavy company where AI features switch on inside approved apps and AI apps receive OAuth access. The view weights where a tool looks at 30 and discovery at 25, because embedded AI and OAuth grants appear in identity and SaaS records. ## The short answer Harmonic Security leads the view "A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access" at 73.5 of 100. Bay is 12th at 50.0. Harmonic Security is ahead mainly on where it looks (7 against 4, worth 9.0 points at weight 30). ## Why these weights Embedded AI and OAuth grants show up in identity and SaaS records, so reading those vantage points weighs most. **Weights used in this view** | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---| | 25 | 30 | 10 | 15 | 10 | 5 | 5 | **A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 25) | Where it looks (weight 30) | Time to a first inventory (weight 10) | What it can do about a finding (weight 15) | Risk context and reporting (weight 10) | Transparency (weight 5) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 73.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | joint 2nd | [Microsoft](/vendors/microsoft/) | 67.5 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | joint 2nd | [Netskope](/vendors/netskope/) | 67.5 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 4th | [Nudge Security](/vendors/nudge-security/) | 64.5 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 5th | [Grip Security](/vendors/grip-security/) | 59.5 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | 6th | [Zscaler](/vendors/zscaler/) | 59.0 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | joint 7th | [Cyberhaven](/vendors/cyberhaven/) | 56.0 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | joint 7th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 56.0 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | 9th | [Reco](/vendors/reco/) | 53.0 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 10th | [Aurascape](/vendors/aurascape/) | 52.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | 11th | [Push Security](/vendors/push-security/) | 51.0 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | 12th | [Bay](/vendors/bay/) | 50.0 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | 13th | [Aim Security](/vendors/aim-security/) | 49.0 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 14th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 48.5 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 15th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 46.5 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 42.5 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## What would change this ranking **Top pair.** Microsoft rising from 7 to 10 on where it looks (weight 30, so each point moves a total by 3.0) would put Microsoft above Harmonic Security; the gap is 6.0 points. **Bay's next place.** Bay rising from 4 to 5 on where it looks (weight 30, so each point moves a total by 3.0) would put Bay above Push Security; the gap is 1.0 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which shadow AI detection tool is best for AI inside approved SaaS? Harmonic Security leads this view. Harmonic says Explore covers AI features embedded in SaaS tools, and its connectors include a compliance API, though none was found on 16 Nov 2026 for discovery of apps granted OAuth access. Microsoft and Netskope share the next position. Microsoft's app governance covers OAuth-enabled apps in Entra ID, Google and Salesforce. Identity-first tools document OAuth discovery directly: Grip Security, Reco, Nudge Security and Push Security each list OAuth-granted AI apps. ## What do endpoint tools miss in this setting? An AI feature switched on inside an approved app produces no new domain and no new install, so endpoint and DNS views see nothing new. Identity, SaaS API and OAuth records are the places it appears. Pair a SaaS API source with a second vantage point, and check how each vendor names embedded features, since a catalog entry for the host app does not show that a new AI setting was enabled. ## Related pages - [All situations](/best-for/) - [An engineering organization where developers add coding agents, IDE plugins and MCP servers on their own](/best-for/developer-heavy/) - [A company that wants staff to keep using AI and be guided to approved tools rather than blocked](/best-for/coach-not-block/) - [Overall ranking](/ranking/) - [Method and weights](/method/) - [Vendor profiles](/vendors/) --- # Best shadow AI detection tool for coaching instead of blocking URL: https://shadowaidetectiontools.com/best-for/coach-not-block/ Updated: 2026-11-16 Harmonic Security leads for a company that wants staff to keep using AI and be guided to approved tools instead of blocked. The view weights what a tool does about a finding at 35, since guiding people in the moment is the job. Weights fixed 16 Nov 2026. ## The short answer Harmonic Security leads the view "A company that wants staff to keep using AI and be guided to approved tools rather than blocked" at 75.5 of 100. Bay is 11th at 53.0. Harmonic Security is ahead mainly on what it can do about a finding (8 against 6, worth 7.0 points at weight 35). ## Why these weights Guiding people in the moment is the job, so what the tool can do when it finds an unapproved app weighs most. **Weights used in this view** | What it discovers | Where it looks | Time to a first inventory | What it can do about a finding | Risk context and reporting | Transparency | Proof | |---|---|---|---|---|---|---| | 20 | 15 | 10 | 35 | 10 | 5 | 5 | **A company that wants staff to keep using AI and be guided to approved tools rather than blocked: scores out of 10 per criterion, totals out of 100** | Rank | Vendor | Total (of 100) | What it discovers (weight 20) | Where it looks (weight 15) | Time to a first inventory (weight 10) | What it can do about a finding (weight 35) | Risk context and reporting (weight 10) | Transparency (weight 5) | Proof (weight 5) | |---|---|---|---|---|---|---|---|---|---| | 1st | [Harmonic Security](/vendors/harmonic-security/) | 75.5 | [7](/vendors/harmonic-security/#evidence-discovers) | [7](/vendors/harmonic-security/#evidence-where) | [8](/vendors/harmonic-security/#evidence-speed) | [8](/vendors/harmonic-security/#evidence-acts) | [9](/vendors/harmonic-security/#evidence-context) | [6](/vendors/harmonic-security/#evidence-transparency) | [6](/vendors/harmonic-security/#evidence-proof) | | joint 2nd | [Microsoft](/vendors/microsoft/) | 65.0 | [8](/vendors/microsoft/#evidence-discovers) | [7](/vendors/microsoft/#evidence-where) | [6](/vendors/microsoft/#evidence-speed) | [6](/vendors/microsoft/#evidence-acts) | [6](/vendors/microsoft/#evidence-context) | [6](/vendors/microsoft/#evidence-transparency) | [5](/vendors/microsoft/#evidence-proof) | | joint 2nd | [Netskope](/vendors/netskope/) | 65.0 | [8](/vendors/netskope/#evidence-discovers) | [7](/vendors/netskope/#evidence-where) | [4](/vendors/netskope/#evidence-speed) | [6](/vendors/netskope/#evidence-acts) | [6](/vendors/netskope/#evidence-context) | [6](/vendors/netskope/#evidence-transparency) | [9](/vendors/netskope/#evidence-proof) | | 4th | [Nudge Security](/vendors/nudge-security/) | 63.0 | [6](/vendors/nudge-security/#evidence-discovers) | [7](/vendors/nudge-security/#evidence-where) | [7](/vendors/nudge-security/#evidence-speed) | [6](/vendors/nudge-security/#evidence-acts) | [6](/vendors/nudge-security/#evidence-context) | [7](/vendors/nudge-security/#evidence-transparency) | [6](/vendors/nudge-security/#evidence-proof) | | 5th | [Cyberhaven](/vendors/cyberhaven/) | 59.5 | [6](/vendors/cyberhaven/#evidence-discovers) | [5](/vendors/cyberhaven/#evidence-where) | [4](/vendors/cyberhaven/#evidence-speed) | [7](/vendors/cyberhaven/#evidence-acts) | [7](/vendors/cyberhaven/#evidence-context) | [3](/vendors/cyberhaven/#evidence-transparency) | [6](/vendors/cyberhaven/#evidence-proof) | | joint 6th | [Aurascape](/vendors/aurascape/) | 58.5 | [7](/vendors/aurascape/#evidence-discovers) | [3](/vendors/aurascape/#evidence-where) | [3](/vendors/aurascape/#evidence-speed) | [7](/vendors/aurascape/#evidence-acts) | [7](/vendors/aurascape/#evidence-context) | [5](/vendors/aurascape/#evidence-transparency) | [6](/vendors/aurascape/#evidence-proof) | | joint 6th | [Grip Security](/vendors/grip-security/) | 58.5 | [5](/vendors/grip-security/#evidence-discovers) | [7](/vendors/grip-security/#evidence-where) | [7](/vendors/grip-security/#evidence-speed) | [6](/vendors/grip-security/#evidence-acts) | [4](/vendors/grip-security/#evidence-context) | [5](/vendors/grip-security/#evidence-transparency) | [7](/vendors/grip-security/#evidence-proof) | | joint 6th | [Zscaler](/vendors/zscaler/) | 58.5 | [7](/vendors/zscaler/#evidence-discovers) | [6](/vendors/zscaler/#evidence-where) | [1](/vendors/zscaler/#evidence-speed) | [6](/vendors/zscaler/#evidence-acts) | [7](/vendors/zscaler/#evidence-context) | [6](/vendors/zscaler/#evidence-transparency) | [7](/vendors/zscaler/#evidence-proof) | | joint 9th | [Palo Alto Networks](/vendors/palo-alto-networks/) | 56.5 | [5](/vendors/palo-alto-networks/#evidence-discovers) | [6](/vendors/palo-alto-networks/#evidence-where) | [5](/vendors/palo-alto-networks/#evidence-speed) | [6](/vendors/palo-alto-networks/#evidence-acts) | [5](/vendors/palo-alto-networks/#evidence-context) | [6](/vendors/palo-alto-networks/#evidence-transparency) | [7](/vendors/palo-alto-networks/#evidence-proof) | | joint 9th | [Push Security](/vendors/push-security/) | 56.5 | [5](/vendors/push-security/#evidence-discovers) | [4](/vendors/push-security/#evidence-where) | [6](/vendors/push-security/#evidence-speed) | [7](/vendors/push-security/#evidence-acts) | [4](/vendors/push-security/#evidence-context) | [7](/vendors/push-security/#evidence-transparency) | [5](/vendors/push-security/#evidence-proof) | | 11th | [Bay](/vendors/bay/) | 53.0 | [6](/vendors/bay/#evidence-discovers) | [4](/vendors/bay/#evidence-where) | [7](/vendors/bay/#evidence-speed) | [6](/vendors/bay/#evidence-acts) | [4](/vendors/bay/#evidence-context) | [0](/vendors/bay/#evidence-transparency) | [6](/vendors/bay/#evidence-proof) | | 12th | [Aim Security](/vendors/aim-security/) | 52.0 | [6](/vendors/aim-security/#evidence-discovers) | [4](/vendors/aim-security/#evidence-where) | [3](/vendors/aim-security/#evidence-speed) | [6](/vendors/aim-security/#evidence-acts) | [5](/vendors/aim-security/#evidence-context) | [2](/vendors/aim-security/#evidence-transparency) | [8](/vendors/aim-security/#evidence-proof) | | 13th | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | 51.5 | [6](/vendors/sentinelone/#evidence-discovers) | [4](/vendors/sentinelone/#evidence-where) | [4](/vendors/sentinelone/#evidence-speed) | [6](/vendors/sentinelone/#evidence-acts) | [4](/vendors/sentinelone/#evidence-context) | [3](/vendors/sentinelone/#evidence-transparency) | [6](/vendors/sentinelone/#evidence-proof) | | 14th | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | 50.0 | [5](/vendors/layerx/#evidence-discovers) | [4](/vendors/layerx/#evidence-where) | [5](/vendors/layerx/#evidence-speed) | [6](/vendors/layerx/#evidence-acts) | [5](/vendors/layerx/#evidence-context) | [1](/vendors/layerx/#evidence-transparency) | [5](/vendors/layerx/#evidence-proof) | | 15th | [Reco](/vendors/reco/) | 46.0 | [5](/vendors/reco/#evidence-discovers) | [7](/vendors/reco/#evidence-where) | [6](/vendors/reco/#evidence-speed) | [3](/vendors/reco/#evidence-acts) | [5](/vendors/reco/#evidence-context) | [1](/vendors/reco/#evidence-transparency) | [7](/vendors/reco/#evidence-proof) | | 16th | [WitnessAI](/vendors/witnessai/) | 45.0 | [6](/vendors/witnessai/#evidence-discovers) | [3](/vendors/witnessai/#evidence-where) | [3](/vendors/witnessai/#evidence-speed) | [5](/vendors/witnessai/#evidence-acts) | [4](/vendors/witnessai/#evidence-context) | [2](/vendors/witnessai/#evidence-transparency) | [6](/vendors/witnessai/#evidence-proof) | ## What would change this ranking **Top pair.** Microsoft rising from 6 to 10 on what it can do about a finding (weight 35, so each point moves a total by 3.5) would put Microsoft above Harmonic Security; the gap is 10.5 points. **Bay's next place.** Bay rising from 6 to 8 on what it can do about a finding (weight 35, so each point moves a total by 3.5) would put Bay above Push Security; the gap is 3.5 points. A score changes only when a sourced fact changes, and the weights were fixed before the scores. The [method page](/method/) lists the weights, anchors and rules. ## Which tool is best when you want to coach rather than block? Harmonic Security leads this view. Harmonic Guide gives inline warn and nudge messages, restricts personal-account use, and records MCP access requests, with block available when a team needs it. Microsoft and Netskope share the next position. Netskope documents coaching, redaction and blocking, and Nudge Security is built around nudges toward approved tools, one-click redaction and an approval status per agent, though a customer on its page says it nudges instead of blocking. ## What should a coaching program measure? Look for the in-the-moment message, an approved alternative to point to, an approval or exception path, and a record of how often users changed course. Push Security documents warnings with alternatives, and Aurascape documents time-limited exception requests. Coaching works best after discovery shows what is in use, so most programs start with the inventory and add warnings per tool. ## Related pages - [All situations](/best-for/) - [A SaaS-heavy company where AI features switch on inside approved apps and AI apps get OAuth access](/best-for/saas-embedded/) - [A regulated company that must show auditors every AI tool touching customer data](/best-for/regulated-inventory/) - [Overall ranking](/ranking/) - [Method and weights](/method/) - [Vendor profiles](/vendors/) --- # Vantage Point Matrix: where shadow AI detection tools look URL: https://shadowaidetectiontools.com/vantage-point-matrix/ Updated: 2026-11-16 The Vantage Point Matrix shows where 16 shadow AI detection tools read from to find unapproved AI: endpoint, browser, network or proxy, identity provider, and SaaS APIs, email and billing. Each cell is yes, partial or none found as of 16 Nov 2026, and cites the fact ids from the vendor's public pages. ## The matrix **Vantage Point Matrix: each vendor against each dimension** | Vendor | [Endpoint](/vantage-point-matrix/endpoint/) | [Browser](/vantage-point-matrix/browser/) | [Network and proxy](/vantage-point-matrix/network-and-proxy/) | [Identity provider](/vantage-point-matrix/identity-provider/) | [SaaS APIs, email and billing](/vantage-point-matrix/saas-apis-email-and-billing/) | |---|---|---|---|---|---| | [Aim Security](/vendors/aim-security/) | [Yes](/vendors/aim-security/#fact-aim-security-11) | [Yes](/vendors/aim-security/#fact-aim-security-16) | [Yes](/vendors/aim-security/#fact-aim-security-16) | [None found](/vendors/aim-security/#fact-aim-security-18) | [None found](/vendors/aim-security/#fact-aim-security-18) | | [Aurascape](/vendors/aurascape/) | [Partial](/vendors/aurascape/#fact-aurascape-15) | [Partial](/vendors/aurascape/#fact-aurascape-15) | [Yes](/vendors/aurascape/#fact-aurascape-12) | [None found](/vendors/aurascape/#fact-aurascape-17) | [Partial](/vendors/aurascape/#fact-aurascape-16) | | [Bay](/vendors/bay/) | [Yes](/vendors/bay/#fact-bay-16) | [Yes](/vendors/bay/#fact-bay-12) | [None found](/vendors/bay/#fact-bay-16) | [Partial](/vendors/bay/#fact-bay-30) | [None found](/vendors/bay/#fact-bay-33) | | [Cyberhaven](/vendors/cyberhaven/) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-01) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-09) | [None found](/vendors/cyberhaven/#fact-cyberhaven-12) | [Partial](/vendors/cyberhaven/#fact-cyberhaven-10) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-08) | | [Grip Security](/vendors/grip-security/) | [None found](/vendors/grip-security/#fact-grip-security-12) | [Yes](/vendors/grip-security/#fact-grip-security-10) | [Partial](/vendors/grip-security/#fact-grip-security-11) | [Yes](/vendors/grip-security/#fact-grip-security-10) | [Yes](/vendors/grip-security/#fact-grip-security-10) | | [Harmonic Security](/vendors/harmonic-security/) | [Yes](/vendors/harmonic-security/#fact-harmonic-security-12) | [Yes](/vendors/harmonic-security/#fact-harmonic-security-15) | [Partial](/vendors/harmonic-security/#fact-harmonic-security-22) | [Partial](/vendors/harmonic-security/#fact-harmonic-security-17) | [Partial](/vendors/harmonic-security/#fact-harmonic-security-19) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | [Partial](/vendors/layerx/#fact-layerx-12) | [Yes](/vendors/layerx/#fact-layerx-12) | [None found](/vendors/layerx/#fact-layerx-16) | [Partial](/vendors/layerx/#fact-layerx-14) | [None found](/vendors/layerx/#fact-layerx-11) | | [Microsoft](/vendors/microsoft/) | [Yes](/vendors/microsoft/#fact-microsoft-02) | [Yes](/vendors/microsoft/#fact-microsoft-08) | [Yes](/vendors/microsoft/#fact-microsoft-02) | [Partial](/vendors/microsoft/#fact-microsoft-06) | [Yes](/vendors/microsoft/#fact-microsoft-01) | | [Netskope](/vendors/netskope/) | [Yes](/vendors/netskope/#fact-netskope-05) | None found | [Yes](/vendors/netskope/#fact-netskope-15) | [Partial](/vendors/netskope/#fact-netskope-11) | [Partial](/vendors/netskope/#fact-netskope-12) | | [Nudge Security](/vendors/nudge-security/) | [None found](/vendors/nudge-security/#fact-nudge-security-10) | [Yes](/vendors/nudge-security/#fact-nudge-security-09) | [None found](/vendors/nudge-security/#fact-nudge-security-10) | [Yes](/vendors/nudge-security/#fact-nudge-security-11) | [Yes](/vendors/nudge-security/#fact-nudge-security-08) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | [None found](/vendors/palo-alto-networks/#fact-palo-alto-networks-13) | [Yes](/vendors/palo-alto-networks/#fact-palo-alto-networks-08) | [Yes](/vendors/palo-alto-networks/#fact-palo-alto-networks-06) | [None found](/vendors/palo-alto-networks/#fact-palo-alto-networks-12) | [Partial](/vendors/palo-alto-networks/#fact-palo-alto-networks-07) | | [Push Security](/vendors/push-security/) | [None found](/vendors/push-security/#fact-push-security-08) | [Yes](/vendors/push-security/#fact-push-security-09) | [None found](/vendors/push-security/#fact-push-security-09) | [Partial](/vendors/push-security/#fact-push-security-10) | [Partial](/vendors/push-security/#fact-push-security-05) | | [Reco](/vendors/reco/) | [None found](/vendors/reco/#fact-reco-09) | [Partial](/vendors/reco/#fact-reco-05) | [Partial](/vendors/reco/#fact-reco-05) | [Yes](/vendors/reco/#fact-reco-06) | [Yes](/vendors/reco/#fact-reco-03) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | [Yes](/vendors/sentinelone/#fact-sentinelone-13) | [Yes](/vendors/sentinelone/#fact-sentinelone-13) | [None found](/vendors/sentinelone/#fact-sentinelone-16) | [None found](/vendors/sentinelone/#fact-sentinelone-16) | [Partial](/vendors/sentinelone/#fact-sentinelone-14) | | [WitnessAI](/vendors/witnessai/) | [None found](/vendors/witnessai/#fact-witnessai-06) | [None found](/vendors/witnessai/#fact-witnessai-06) | [Yes](/vendors/witnessai/#fact-witnessai-04) | [None found](/vendors/witnessai/#fact-witnessai-10) | [None found](/vendors/witnessai/#fact-witnessai-10) | | [Zscaler](/vendors/zscaler/) | [Yes](/vendors/zscaler/#fact-zscaler-11) | [Partial](/vendors/zscaler/#fact-zscaler-06) | [Yes](/vendors/zscaler/#fact-zscaler-07) | [None found](/vendors/zscaler/#fact-zscaler-26) | [None found](/vendors/zscaler/#fact-zscaler-13) | - **Yes**: documented specifically - **Partial**: documented in general terms, for some tools only, or as an alert without the action - **None found**: none found on 16 Nov 2026 ## Why the vantage point decides what you find A detection tool only finds what its sensors can see. A proxy sees traffic and no installed files, an endpoint sensor sees files and no personal phone, and an identity log sees work sign-ins and no private accounts. Read the matrix by vantage point first, then by vendor. A gap on one row can be closed by a vendor strong on another, and the per-dimension pages explain what each row sees and misses. ## Download and cite [Download the Vantage Point Matrix as CSV or JSON](/datasets/vantage-point-matrix/). Cite it as: Shadow AI Detection Tools, Vantage Point Matrix and Finding-to-Action Ladder, edition of 16 November 2026, https://shadowaidetectiontools.com/, licensed CC BY 4.0. ## Related pages - [How the Vantage Point Matrix is built](/vantage-point-matrix/methodology/) - [Vantage Point Matrix data download](/datasets/vantage-point-matrix/) - [Endpoint](/vantage-point-matrix/endpoint/) - [Browser](/vantage-point-matrix/browser/) - [Network and proxy](/vantage-point-matrix/network-and-proxy/) - [Identity provider](/vantage-point-matrix/identity-provider/) - [SaaS APIs, email and billing](/vantage-point-matrix/saas-apis-email-and-billing/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) - [Vendor profiles](/vendors/) - [The ranking](/ranking/) --- # Vantage Point Matrix methodology URL: https://shadowaidetectiontools.com/vantage-point-matrix/methodology/ Updated: 2026-11-16 The Vantage Point Matrix records, for 16 shadow AI detection vendors, whether each documents a given vantage point as a discovery source: endpoint, browser, network or proxy, identity provider, and SaaS APIs, email and billing. Every cell cites fact ids from public vendor pages read on 16 Nov 2026, and the rule is the same for every vendor. ## What the matrix is The matrix lists 16 vendors against five vantage points, the places a tool reads from to find unapproved AI. A cell says whether the vendor's public pages document that vantage point as a source of discovery, and cites the fact ids that support the call. Every cell is copied unchanged from the matrix recorded in that vendor's evidence file during the 16 Nov 2026 evidence pass (the shadowaidetectiontools.vantage matrix), so the levels, fact ids and notes are the ones the evidence pass wrote. ## The rule Yes means a fact names the vantage point as a discovery source specifically. Partial means it is named in general terms, covers some uses only, or is used for policy context and not documented as discovery. None found means no fact in the vendor's evidence file names it on 16 Nov 2026. The same rule applies to every vendor, Bay included. A vendor's own claim is treated as a claim: it is recorded as documented, not as tested. ## Sources Cells come from the vendor evidence files for each vendor: product pages, documentation, release notes and press releases, plus third-party reports where cited. Each fact carries its URL and read date, and the cells cite the fact ids. ## Read date All pages were read on 16 Nov 2026. A change after that date is not reflected until the matrix is next updated. ## How to cite Cite as: Shadow AI Detection Tools, Vantage Point Matrix, edition of 16 Nov 2026, https://shadowaidetectiontools.com/vantage-point-matrix/, CC BY 4.0. ## License The matrix and its CSV and JSON files are published under CC BY 4.0. Reuse with the citation above. ## Corrections To correct a cell, send the vendor, the dimension and a public page that shows the documented source. Accepted changes are logged on the corrections page with their date. ## Levels and counts - **Yes** (34 cells): documented specifically - **Partial** (21 cells): documented in general terms, for some tools only, or as an alert without the action - **None found** (25 cells): none found on 16 Nov 2026 ## Related pages - [The Vantage Point Matrix](/vantage-point-matrix/) - [Vantage Point Matrix download](/datasets/vantage-point-matrix/) - [Scoring method](/method/) - [Endpoint](/vantage-point-matrix/endpoint/) - [Browser](/vantage-point-matrix/browser/) - [Network and proxy](/vantage-point-matrix/network-and-proxy/) --- # Endpoint vantage point for shadow AI detection tools URL: https://shadowaidetectiontools.com/vantage-point-matrix/endpoint/ Updated: 2026-11-16 The endpoint vantage point is a sensor or script on the device that lists AI apps, agents, extensions and MCP servers installed there. Across the 16 vendors read on 16 Nov 2026, 8 document it specifically, 2 in general terms or for some uses, and 6 show none found. Each vendor's cell below cites the fact ids behind it. ## Endpoint: every vendor **Endpoint by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | Yes | Lightweight endpoint client finds agents and AI apps. | [aim-security-11](/vendors/aim-security/#fact-aim-security-11), [aim-security-12](/vendors/aim-security/#fact-aim-security-12) | | [Bay](/vendors/bay/) | Yes | Scans devices via MDM or EDR delivered script and ephemeral binary. | [bay-16](/vendors/bay/#fact-bay-16), [bay-17](/vendors/bay/#fact-bay-17) | | [Cyberhaven](/vendors/cyberhaven/) | Yes | Endpoint agent is the primary sensor. | [cyberhaven-01](/vendors/cyberhaven/#fact-cyberhaven-01), [cyberhaven-04](/vendors/cyberhaven/#fact-cyberhaven-04) | | [Harmonic Security](/vendors/harmonic-security/) | Yes | Endpoint Agent on Windows, macOS and Linux. | [harmonic-security-12](/vendors/harmonic-security/#fact-harmonic-security-12), [harmonic-security-20](/vendors/harmonic-security/#fact-harmonic-security-20) | | [Microsoft](/vendors/microsoft/) | Yes | Defender for Endpoint feeds cloud discovery and agent discovery. | [microsoft-02](/vendors/microsoft/#fact-microsoft-02), [microsoft-11](/vendors/microsoft/#fact-microsoft-11) | | [Netskope](/vendors/netskope/) | Yes | Netskope Client endpoint AI discovery, version 138 or later. | [netskope-05](/vendors/netskope/#fact-netskope-05), [netskope-10](/vendors/netskope/#fact-netskope-10) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Yes | Dedicated AI endpoint agent. | [sentinelone-13](/vendors/sentinelone/#fact-sentinelone-13) | | [Zscaler](/vendors/zscaler/) | Yes | Endpoint AI Security covers managed devices. | [zscaler-11](/vendors/zscaler/#fact-zscaler-11), [zscaler-15](/vendors/zscaler/#fact-zscaler-15) | | [Aurascape](/vendors/aurascape/) | Partial | Local agent discovery and active devices are mentioned; no endpoint agent documented. | [aurascape-15](/vendors/aurascape/#fact-aurascape-15), [aurascape-30](/vendors/aurascape/#fact-aurascape-30) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Partial | Desktop visibility is named; the mechanism is not described. | [layerx-12](/vendors/layerx/#fact-layerx-12), [layerx-13](/vendors/layerx/#fact-layerx-13) | | [Grip Security](/vendors/grip-security/) | None found | No endpoint agent documented; Okta page says no agents or proxies. | [grip-security-12](/vendors/grip-security/#fact-grip-security-12) | | [Nudge Security](/vendors/nudge-security/) | None found | Says it does not touch endpoints. | [nudge-security-10](/vendors/nudge-security/#fact-nudge-security-10), [nudge-security-04](/vendors/nudge-security/#fact-nudge-security-04) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | None found | None found on 2026-11-16; AI Access Security needs no endpoint agent. | [palo-alto-networks-13](/vendors/palo-alto-networks/#fact-palo-alto-networks-13) | | [Push Security](/vendors/push-security/) | None found | None found on 2026-11-16; Push is a browser extension. | [push-security-08](/vendors/push-security/#fact-push-security-08) | | [Reco](/vendors/reco/) | None found | No endpoint agent documented. | [reco-09](/vendors/reco/#fact-reco-09) | | [WitnessAI](/vendors/witnessai/) | None found | Vendor says it needs no endpoint client; no endpoint sensor documented. | [witnessai-06](/vendors/witnessai/#fact-witnessai-06), [witnessai-10](/vendors/witnessai/#fact-witnessai-10) | ## What the endpoint vantage point sees An endpoint sensor reads the device itself, so it can find desktop AI apps, command-line agents, IDE plugins, browser extensions and MCP server configuration that never cross a web proxy. Microsoft says local AI agent discovery is generally available on Windows and macOS and reports each agent's configured MCP servers (October 2026 release notes). Zscaler says Endpoint AI Security identifies AI assistants, agents, development tools, models, packages and extensions on managed devices (5 October 2026). Setup can be light. Bay says it deploys with one script through an existing MDM or EDR, using an ephemeral binary and no proxy. Sources: [Bay source (bay-17)](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=asset-dim); [Zscaler source (zscaler-15)](https://www.zscaler.com/blogs/product-insights/bringing-anthropic-s-claude-mythos-5-1-zscaler-endpoint-ai-security); [Microsoft source (microsoft-15)](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint). ## What the endpoint vantage point misses A device scan covers only devices that carry the sensor or receive the script. Personal phones and unmanaged laptops fall outside it, and AI features switched on inside an approved SaaS application never appear as files on a disk. Tools whose matrix cell reads none found on 16 Nov 2026, such as Nudge Security (which says it does not touch endpoints) and Push Security (a browser extension), rely on other vantage points for this layer. Sources: [Microsoft source (microsoft-15)](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint); [Nudge Security source (nudge-security-10)](https://www.nudgesecurity.com/product). ## Vendors by level Documented specifically (8): Aim Security, Bay, Cyberhaven, Harmonic Security, Microsoft, Netskope, SentinelOne and Zscaler. General terms or partial (2): Aurascape and LayerX. None found on 16 Nov 2026 (6): Grip Security, Nudge Security, Palo Alto Networks, Push Security, Reco and WitnessAI. ## Related pages - [The full Vantage Point Matrix](/vantage-point-matrix/) - [SaaS APIs, email and billing](/vantage-point-matrix/saas-apis-email-and-billing/) - [Browser](/vantage-point-matrix/browser/) - [Method](/vantage-point-matrix/methodology/) - [Vendor profiles](/vendors/) --- # Browser vantage point for shadow AI detection tools URL: https://shadowaidetectiontools.com/vantage-point-matrix/browser/ Updated: 2026-11-16 The browser vantage point is an extension that sees which AI web apps and accounts employees use, including personal accounts. Across the 16 vendors read on 16 Nov 2026, 11 document it specifically, 3 in general terms or for some uses, and 2 show none found. Each vendor's cell below cites the fact ids behind it. ## Browser: every vendor **Browser by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | Yes | Browser AI is inspected and Cato AI extensions are in use. | [aim-security-16](/vendors/aim-security/#fact-aim-security-16), [aim-security-34](/vendors/aim-security/#fact-aim-security-34) | | [Bay](/vendors/bay/) | Yes | Browser coverage with account-level detection, including personal Chrome profiles. | [bay-12](/vendors/bay/#fact-bay-12), [bay-22](/vendors/bay/#fact-bay-22) | | [Cyberhaven](/vendors/cyberhaven/) | Yes | Inspects AI tools in the browser; standalone extension for unmanaged devices. | [cyberhaven-09](/vendors/cyberhaven/#fact-cyberhaven-09), [cyberhaven-13](/vendors/cyberhaven/#fact-cyberhaven-13) | | [Grip Security](/vendors/grip-security/) | Yes | Optional Grip Extend browser extension. | [grip-security-10](/vendors/grip-security/#fact-grip-security-10), [grip-security-13](/vendors/grip-security/#fact-grip-security-13) | | [Harmonic Security](/vendors/harmonic-security/) | Yes | Browser extension across Chromium browsers and Safari. | [harmonic-security-15](/vendors/harmonic-security/#fact-harmonic-security-15) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Yes | Browser extension on Chrome, Edge, Firefox and Safari. | [layerx-12](/vendors/layerx/#fact-layerx-12), [layerx-16](/vendors/layerx/#fact-layerx-16), [layerx-17](/vendors/layerx/#fact-layerx-17) | | [Microsoft](/vendors/microsoft/) | Yes | Purview browser extension is a prerequisite for third-party AI sites. | [microsoft-08](/vendors/microsoft/#fact-microsoft-08) | | [Nudge Security](/vendors/nudge-security/) | Yes | Browser extension on Mac and Windows. | [nudge-security-09](/vendors/nudge-security/#fact-nudge-security-09), [nudge-security-15](/vendors/nudge-security/#fact-nudge-security-15) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Yes | Prisma Browser is a listed enforcement point with extension and GenAI app visibility. | [palo-alto-networks-08](/vendors/palo-alto-networks/#fact-palo-alto-networks-08), [palo-alto-networks-09](/vendors/palo-alto-networks/#fact-palo-alto-networks-09), [palo-alto-networks-10](/vendors/palo-alto-networks/#fact-palo-alto-networks-10) | | [Push Security](/vendors/push-security/) | Yes | Lightweight extension in existing browsers. | [push-security-09](/vendors/push-security/#fact-push-security-09) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Yes | Browser extension for web AI. | [sentinelone-13](/vendors/sentinelone/#fact-sentinelone-13), [sentinelone-11](/vendors/sentinelone/#fact-sentinelone-11) | | [Aurascape](/vendors/aurascape/) | Partial | Browser traffic is decoded; no browser extension documented. | [aurascape-15](/vendors/aurascape/#fact-aurascape-15) | | [Reco](/vendors/reco/) | Partial | Lists browser behavior and extension detection, mechanism not documented. | [reco-05](/vendors/reco/#fact-reco-05) | | [Zscaler](/vendors/zscaler/) | Partial | Browser extension and plugin visibility claimed; isolated browser for controls. | [zscaler-06](/vendors/zscaler/#fact-zscaler-06), [zscaler-19](/vendors/zscaler/#fact-zscaler-19) | | [Netskope](/vendors/netskope/) | None found | None found on 2026-11-16 for a browser extension as a discovery source; browser extensions are discovered on endpoints. | none | | [WitnessAI](/vendors/witnessai/) | None found | Vendor says it works without browser extensions. | [witnessai-06](/vendors/witnessai/#fact-witnessai-06), [witnessai-10](/vendors/witnessai/#fact-witnessai-10) | ## What the browser vantage point sees A browser extension sees the page the employee is using: which AI web app, which account, and what is typed or pasted. Bay says it identifies personal versus corporate AI accounts, including in personal Chrome profiles. Harmonic says its extension covers all Chromium-based browsers and Safari, including AI browsers. Push Security says a lightweight extension in existing browsers sees browser-layer activity that network and endpoint tools lack. Akamai says LayerX needs no proxies, network redesign or browser replacement. Sources: [Bay source (bay-22)](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=asset-dim); [Harmonic Security source (harmonic-security-15)](https://www.harmonic.security/resources/chatgpt-atlas-protecting-employee-use-of-ai-browsers); [Push Security source (push-security-09)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## What the browser vantage point misses An extension sees only browsers where it is installed and running. Desktop AI apps, terminal agents and background services do not pass through it, and a user on an unmanaged browser profile is invisible unless the vendor documents a way in. WitnessAI says it works without browser extensions, so its cell reads none found on 16 Nov 2026 for this vantage point. Sources: [Push Security source (push-security-09)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [LayerX (Akamai Workforce Protector) source (layerx-16)](https://www.akamai.com/products/akamai-workforce-protector). ## Vendors by level Documented specifically (11): Aim Security, Bay, Cyberhaven, Grip Security, Harmonic Security, LayerX, Microsoft, Nudge Security, Palo Alto Networks, Push Security and SentinelOne. General terms or partial (3): Aurascape, Reco and Zscaler. None found on 16 Nov 2026 (2): Netskope and WitnessAI. ## Related pages - [The full Vantage Point Matrix](/vantage-point-matrix/) - [Endpoint](/vantage-point-matrix/endpoint/) - [Network and proxy](/vantage-point-matrix/network-and-proxy/) - [Method](/vantage-point-matrix/methodology/) - [Vendor profiles](/vendors/) --- # Network and proxy vantage point for shadow AI detection URL: https://shadowaidetectiontools.com/vantage-point-matrix/network-and-proxy/ Updated: 2026-11-16 The network and proxy vantage point reads traffic through a gateway, firewall or log collector to find AI web apps and APIs. Across the 16 vendors read on 16 Nov 2026, 7 document it specifically, 3 in general terms or for some uses, and 6 show none found. Each vendor's cell below cites the fact ids behind it. ## Network and proxy: every vendor **Network and proxy by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | Yes | Browser and app AI traffic is inspected inline on the network, agentless. | [aim-security-16](/vendors/aim-security/#fact-aim-security-16), [aim-security-17](/vendors/aim-security/#fact-aim-security-17) | | [Aurascape](/vendors/aurascape/) | Yes | Inline AI proxy inspects AI traffic and steers it alongside existing SASE. | [aurascape-12](/vendors/aurascape/#fact-aurascape-12), [aurascape-13](/vendors/aurascape/#fact-aurascape-13), [aurascape-14](/vendors/aurascape/#fact-aurascape-14) | | [Microsoft](/vendors/microsoft/) | Yes | Firewall and proxy log collector, Zscaler integration, SSE integration. | [microsoft-02](/vendors/microsoft/#fact-microsoft-02), [microsoft-09](/vendors/microsoft/#fact-microsoft-09) | | [Netskope](/vendors/netskope/) | Yes | Next-Gen Secure Web Gateway inline traffic is the source for AI app discovery. | [netskope-15](/vendors/netskope/#fact-netskope-15), [netskope-04](/vendors/netskope/#fact-netskope-04) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Yes | Reads network traffic through NGFW and Prisma Access. | [palo-alto-networks-06](/vendors/palo-alto-networks/#fact-palo-alto-networks-06), [palo-alto-networks-10](/vendors/palo-alto-networks/#fact-palo-alto-networks-10) | | [WitnessAI](/vendors/witnessai/) | Yes | Network-level discovery, with an F5 BIG-IP SSL Orchestrator integration for encrypted traffic. | [witnessai-04](/vendors/witnessai/#fact-witnessai-04), [witnessai-06](/vendors/witnessai/#fact-witnessai-06), [witnessai-09](/vendors/witnessai/#fact-witnessai-09) | | [Zscaler](/vendors/zscaler/) | Yes | Inline proxy path across AI traffic. | [zscaler-07](/vendors/zscaler/#fact-zscaler-07), [zscaler-12](/vendors/zscaler/#fact-zscaler-12) | | [Grip Security](/vendors/grip-security/) | Partial | Cisco Umbrella integration lists blocking of discovered apps; no traffic discovery documented. | [grip-security-11](/vendors/grip-security/#fact-grip-security-11), [grip-security-19](/vendors/grip-security/#fact-grip-security-19) | | [Harmonic Security](/vendors/harmonic-security/) | Partial | Reports connections to known AI hosts from the endpoint; no proxy or firewall integration documented. | [harmonic-security-22](/vendors/harmonic-security/#fact-harmonic-security-22), [harmonic-security-21](/vendors/harmonic-security/#fact-harmonic-security-21) | | [Reco](/vendors/reco/) | Partial | Lists network traffic analysis in general terms. | [reco-05](/vendors/reco/#fact-reco-05) | | [Bay](/vendors/bay/) | None found | Vendor states no proxy; no network, DNS or firewall vantage found. | [bay-16](/vendors/bay/#fact-bay-16), [bay-33](/vendors/bay/#fact-bay-33) | | [Cyberhaven](/vendors/cyberhaven/) | None found | None found on 2026-11-16 for proxy, firewall or DNS vantage; the page says network-layer tools miss OS-level agents. | [cyberhaven-12](/vendors/cyberhaven/#fact-cyberhaven-12) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | None found | Network inspection is explicitly not used; none found as a vantage point. | [layerx-16](/vendors/layerx/#fact-layerx-16) | | [Nudge Security](/vendors/nudge-security/) | None found | No proxy, firewall or DNS vantage point; says it does not touch the network. | [nudge-security-10](/vendors/nudge-security/#fact-nudge-security-10) | | [Push Security](/vendors/push-security/) | None found | None found on 2026-11-16; Push contrasts itself with network tools. | [push-security-09](/vendors/push-security/#fact-push-security-09) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | None found | None found as a discovery source; vendor says network controls alone fall short. | [sentinelone-16](/vendors/sentinelone/#fact-sentinelone-16) | ## What the network and proxy vantage point sees Traffic inspection finds AI web apps and APIs used from any device that routes through the gateway, without installing anything on the device. Aurascape says its discovery inspects and categorizes AI traffic in motion. WitnessAI says it works over network connectivity and names an integration with F5 BIG-IP SSL Orchestrator to see encrypted AI traffic. Microsoft documents three cloud discovery feeds, one of which is a log collector on firewalls and proxies, plus proxy integrations such as Zscaler. Netskope docs say AI application discovery requires its Next-Gen Secure Web Gateway to be enabled, so setup depends on an existing deployment. Sources: [Aurascape source (aurascape-12)](https://aurascape.ai/discover-and-monitor-ai/); [WitnessAI source (witnessai-09)](https://witness.ai/resources/witnessai-announces-integration-with-f5-big-ip-to-deliver-visibility-and-control-over-encrypted-genai-activity/); [Microsoft source (microsoft-02)](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it). ## What the network and proxy vantage point misses Traffic that never reaches the gateway is missed: a laptop off the corporate network, a personal hotspot, or local agents that talk to a model through an API key. Encrypted traffic needs inspection or is seen only as a domain. Nudge Security says it does not touch the corporate network, and Bay says it uses no proxy, so both cells read none found on 16 Nov 2026. Sources: [Microsoft source (microsoft-02)](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Netskope source (netskope-15)](https://docs.netskope.com/en/ai-command-center); [Nudge Security source (nudge-security-10)](https://www.nudgesecurity.com/product). ## Vendors by level Documented specifically (7): Aim Security, Aurascape, Microsoft, Netskope, Palo Alto Networks, WitnessAI and Zscaler. General terms or partial (3): Grip Security, Harmonic Security and Reco. None found on 16 Nov 2026 (6): Bay, Cyberhaven, LayerX, Nudge Security, Push Security and SentinelOne. ## Related pages - [The full Vantage Point Matrix](/vantage-point-matrix/) - [Browser](/vantage-point-matrix/browser/) - [Identity provider](/vantage-point-matrix/identity-provider/) - [Method](/vantage-point-matrix/methodology/) - [Vendor profiles](/vendors/) --- # Identity provider vantage point for shadow AI detection URL: https://shadowaidetectiontools.com/vantage-point-matrix/identity-provider/ Updated: 2026-11-16 The identity provider vantage point reads sign-on logs, app registrations and OAuth data to find AI services used with a work identity. Across the 16 vendors read on 16 Nov 2026, 3 document it specifically, 7 in general terms or for some uses, and 6 show none found. Each vendor's cell below cites the fact ids behind it. ## Identity provider: every vendor **Identity provider by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Grip Security](/vendors/grip-security/) | Yes | Identity provider connection plus Okta and Entra ID connectors. | [grip-security-10](/vendors/grip-security/#fact-grip-security-10), [grip-security-11](/vendors/grip-security/#fact-grip-security-11) | | [Nudge Security](/vendors/nudge-security/) | Yes | Microsoft 365, Google Workspace, Okta and Entra ID. | [nudge-security-11](/vendors/nudge-security/#fact-nudge-security-11) | | [Reco](/vendors/reco/) | Yes | Reads SSO logs; Okta and Microsoft 365 listed as integrations. | [reco-06](/vendors/reco/#fact-reco-06), [reco-07](/vendors/reco/#fact-reco-07) | | [Bay](/vendors/bay/) | Partial | Connects to the identity provider to scope policy; use as a discovery source not documented. | [bay-30](/vendors/bay/#fact-bay-30) | | [Cyberhaven](/vendors/cyberhaven/) | Partial | Uses Okta, Entra ID and Google Workspace for policy and incident context, not documented as an AI discovery source. | [cyberhaven-10](/vendors/cyberhaven/#fact-cyberhaven-10) | | [Harmonic Security](/vendors/harmonic-security/) | Partial | Entra, Okta and Google enrich employee identity; none found on 2026-11-16 for discovering OAuth grants. | [harmonic-security-17](/vendors/harmonic-security/#fact-harmonic-security-17), [harmonic-security-18](/vendors/harmonic-security/#fact-harmonic-security-18) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Partial | Integrates with IAM and identity providers, no products named. | [layerx-14](/vendors/layerx/#fact-layerx-14) | | [Microsoft](/vendors/microsoft/) | Partial | Entra app registrations through app governance. | [microsoft-06](/vendors/microsoft/#fact-microsoft-06) | | [Netskope](/vendors/netskope/) | Partial | Known identities come from the identity provider; no IdP log discovery described. | [netskope-11](/vendors/netskope/#fact-netskope-11) | | [Push Security](/vendors/push-security/) | Partial | Integrates with M365, Google Workspace or Okta to sync employees. | [push-security-10](/vendors/push-security/#fact-push-security-10) | | [Aim Security](/vendors/aim-security/) | None found | None found on 2026-11-16. | [aim-security-18](/vendors/aim-security/#fact-aim-security-18) | | [Aurascape](/vendors/aurascape/) | None found | No named identity provider integration found. | [aurascape-17](/vendors/aurascape/#fact-aurascape-17) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | None found | None found on 2026-11-16. | [palo-alto-networks-12](/vendors/palo-alto-networks/#fact-palo-alto-networks-12) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | None found | None found on 2026-11-16. | [sentinelone-16](/vendors/sentinelone/#fact-sentinelone-16) | | [WitnessAI](/vendors/witnessai/) | None found | No Okta, Entra ID or Google discovery source found. | [witnessai-10](/vendors/witnessai/#fact-witnessai-10) | | [Zscaler](/vendors/zscaler/) | None found | No identity provider source found. | [zscaler-26](/vendors/zscaler/#fact-zscaler-26) | ## What the identity provider vantage point sees Sign-on logs and app registrations show which AI services employees sign in to with a work identity, with no agent on any device. Reco says Application Discovery draws on SSO logs, email metadata and OAuth integration monitoring. Nudge Security says it integrates with Microsoft 365, Google Workspace, Okta and Entra ID. Grip says discovery connects the customer's email and identity provider. Identity data is also used only for context. Cyberhaven says it pulls directory context from Okta, Entra ID and Google Workspace for policies and incident review, which the matrix records as partial because it is not documented as a discovery source. Sources: [Reco source (reco-06)](https://www.reco.ai/platform/application-discovery); [Nudge Security source (nudge-security-11)](https://www.nudgesecurity.com/use-cases/ai-security-3); [Grip Security source (grip-security-10)](https://www.grip.security/platform/discover-saas-risks). ## What the identity provider vantage point misses Identity logs miss anything an employee reaches without single sign-on, such as a personal account created with a private email address, and tools used without any login. Several vendors, including WitnessAI and Zscaler, show none found on 16 Nov 2026 for an identity provider source. Sources: [Grip Security source (grip-security-10)](https://www.grip.security/platform/discover-saas-risks); [Cyberhaven source (cyberhaven-10)](https://www.cyberhaven.com/product/integrations). ## Vendors by level Documented specifically (3): Grip Security, Nudge Security and Reco. General terms or partial (7): Bay, Cyberhaven, Harmonic Security, LayerX, Microsoft, Netskope and Push Security. None found on 16 Nov 2026 (6): Aim Security, Aurascape, Palo Alto Networks, SentinelOne, WitnessAI and Zscaler. ## Related pages - [The full Vantage Point Matrix](/vantage-point-matrix/) - [Network and proxy](/vantage-point-matrix/network-and-proxy/) - [SaaS APIs, email and billing](/vantage-point-matrix/saas-apis-email-and-billing/) - [Method](/vantage-point-matrix/methodology/) - [Vendor profiles](/vendors/) --- # SaaS APIs, email and billing records for shadow AI detection URL: https://shadowaidetectiontools.com/vantage-point-matrix/saas-apis-email-and-billing/ Updated: 2026-11-16 The SaaS APIs, email and billing vantage point reads connected tenants, mailboxes and records to find AI tools staff signed up for. Across the 16 vendors read on 16 Nov 2026, 5 document it specifically, 6 in general terms or for some uses, and 5 show none found. Each vendor's cell below cites the fact ids behind it. ## SaaS APIs, email and billing: every vendor **SaaS APIs, email and billing by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Cyberhaven](/vendors/cyberhaven/) | Yes | Reads ChatGPT Enterprise and Claude Enterprise through Compliance APIs and connects SaaS and cloud data stores. | [cyberhaven-08](/vendors/cyberhaven/#fact-cyberhaven-08), [cyberhaven-09](/vendors/cyberhaven/#fact-cyberhaven-09) | | [Grip Security](/vendors/grip-security/) | Yes | Email connection, read-only APIs and 70+ integrations. | [grip-security-10](/vendors/grip-security/#fact-grip-security-10), [grip-security-15](/vendors/grip-security/#fact-grip-security-15) | | [Microsoft](/vendors/microsoft/) | Yes | App connectors and app governance read SaaS data. | [microsoft-01](/vendors/microsoft/#fact-microsoft-01), [microsoft-06](/vendors/microsoft/#fact-microsoft-06) | | [Nudge Security](/vendors/nudge-security/) | Yes | Email-based discovery plus connected SaaS app APIs. | [nudge-security-08](/vendors/nudge-security/#fact-nudge-security-08), [nudge-security-11](/vendors/nudge-security/#fact-nudge-security-11) | | [Reco](/vendors/reco/) | Yes | Reads OAuth, email metadata and SaaS integrations (285+ app integrations claimed). | [reco-03](/vendors/reco/#fact-reco-03), [reco-06](/vendors/reco/#fact-reco-06) | | [Aurascape](/vendors/aurascape/) | Partial | Connects to file repositories for Copilot readiness; no SaaS API discovery of AI apps documented. | [aurascape-16](/vendors/aurascape/#fact-aurascape-16) | | [Harmonic Security](/vendors/harmonic-security/) | Partial | Anthropic Compliance API and OpenTelemetry feeds documented; none found on 2026-11-16 for email or billing records. | [harmonic-security-19](/vendors/harmonic-security/#fact-harmonic-security-19) | | [Netskope](/vendors/netskope/) | Partial | API-based protection mentioned for AI apps; API-based discovery not detailed. | [netskope-12](/vendors/netskope/#fact-netskope-12) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Partial | Plugin and SaaS posture data come from SaaS Security Posture Management; named SaaS integrations not read. | [palo-alto-networks-07](/vendors/palo-alto-networks/#fact-palo-alto-networks-07) | | [Push Security](/vendors/push-security/) | Partial | OAuth grant discovery and SIEM output; no SaaS audit-log or billing source found. | [push-security-05](/vendors/push-security/#fact-push-security-05), [push-security-11](/vendors/push-security/#fact-push-security-11) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Partial | API integrations are named in general terms; no SaaS logs or billing sources found. | [sentinelone-14](/vendors/sentinelone/#fact-sentinelone-14), [sentinelone-15](/vendors/sentinelone/#fact-sentinelone-15) | | [Aim Security](/vendors/aim-security/) | None found | None found on 2026-11-16. | [aim-security-18](/vendors/aim-security/#fact-aim-security-18) | | [Bay](/vendors/bay/) | None found | No SaaS API, audit log, email or billing records vantage found. | [bay-33](/vendors/bay/#fact-bay-33) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | None found | None found on 2026-11-16. | [layerx-11](/vendors/layerx/#fact-layerx-11) | | [WitnessAI](/vendors/witnessai/) | None found | No SaaS API, audit log, email or billing discovery found. | [witnessai-10](/vendors/witnessai/#fact-witnessai-10) | | [Zscaler](/vendors/zscaler/) | None found | No SaaS API, email or billing source found for shadow AI; cloud APIs used for MCP and model assets. | [zscaler-13](/vendors/zscaler/#fact-zscaler-13), [zscaler-26](/vendors/zscaler/#fact-zscaler-26) | ## What the saas apis, email and billing vantage point sees API reads find AI that is already inside a connected SaaS tenant. Cyberhaven says it reads ChatGPT Enterprise and Claude Enterprise through their Compliance APIs (23 July 2026). Reco says Application Discovery monitors OAuth apps, browser extensions and APIs. Push Security says it identifies OAuth connections between AI tools and corporate systems such as Google Workspace. Email analysis finds sign-up and receipt messages. Nudge Security says discovery starts from email analysis through a Microsoft 365 or Google Workspace integration, then adds a browser extension and SaaS API integrations. Grip says its free Shadow AI Analysis sets up in about 10 minutes through read-only APIs. Sources: [Cyberhaven source (cyberhaven-08)](https://www.cyberhaven.com/press-releases/cyberhaven-extends-ai-security-chatgpt-enterprise-claude); [Reco source (reco-03)](https://www.reco.ai/platform/application-discovery); [Push Security source (push-security-05)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## What the saas apis, email and billing vantage point misses API reads reach only the tenants an administrator connects, and email analysis sees only mailboxes in scope. Billing and expense records are the third source in this dimension: no claim in the 16 vendors' evidence files documents them as a discovery source on 16 Nov 2026. Local tools that need no account, such as a model run on a laptop, leave no trace in any SaaS API. Sources: [Push Security source (push-security-05)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [Nudge Security source (nudge-security-08)](https://www.nudgesecurity.com/use-cases/ai-security-3); [Grip Security source (grip-security-15)](https://www.grip.security/pricing). ## Vendors by level Documented specifically (5): Cyberhaven, Grip Security, Microsoft, Nudge Security and Reco. General terms or partial (6): Aurascape, Harmonic Security, Netskope, Palo Alto Networks, Push Security and SentinelOne. None found on 16 Nov 2026 (5): Aim Security, Bay, LayerX, WitnessAI and Zscaler. ## Related pages - [The full Vantage Point Matrix](/vantage-point-matrix/) - [Identity provider](/vantage-point-matrix/identity-provider/) - [Endpoint](/vantage-point-matrix/endpoint/) - [Method](/vantage-point-matrix/methodology/) - [Vendor profiles](/vendors/) --- # Finding-to-Action Ladder: what tools do after finding shadow AI URL: https://shadowaidetectiontools.com/finding-to-action-ladder/ Updated: 2026-11-16 The Finding-to-Action Ladder shows what 16 shadow AI detection tools document doing once they find an unapproved AI tool: warn or coach, block, require approval, revoke OAuth grants, redact or stop data, and route to a ticket. Cells read yes, partial or none found as of 16 Nov 2026 and cite fact ids. ## The matrix **Finding-to-Action Ladder: each vendor against each dimension** | Vendor | [Warn or coach](/finding-to-action-ladder/warn-or-coach/) | [Block](/finding-to-action-ladder/block/) | [Request approval](/finding-to-action-ladder/request-approval/) | [Revoke OAuth grant](/finding-to-action-ladder/revoke-oauth-grant/) | [Redact or stop data](/finding-to-action-ladder/redact-or-stop-data/) | [Route to ticket](/finding-to-action-ladder/route-to-ticket/) | |---|---|---|---|---|---|---| | [Aim Security](/vendors/aim-security/) | [Partial](/vendors/aim-security/#fact-aim-security-23) | [Yes](/vendors/aim-security/#fact-aim-security-24) | None found | None found | [Yes](/vendors/aim-security/#fact-aim-security-22) | None found | | [Aurascape](/vendors/aurascape/) | [Yes](/vendors/aurascape/#fact-aurascape-23) | [Yes](/vendors/aurascape/#fact-aurascape-21) | [Partial](/vendors/aurascape/#fact-aurascape-22) | [None found](/vendors/aurascape/#fact-aurascape-26) | [Yes](/vendors/aurascape/#fact-aurascape-23) | [None found](/vendors/aurascape/#fact-aurascape-26) | | [Bay](/vendors/bay/) | [Yes](/vendors/bay/#fact-bay-23) | [Yes](/vendors/bay/#fact-bay-24) | [Partial](/vendors/bay/#fact-bay-25) | [None found](/vendors/bay/#fact-bay-32) | [Yes](/vendors/bay/#fact-bay-23) | None found | | [Cyberhaven](/vendors/cyberhaven/) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-16) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-16) | [None found](/vendors/cyberhaven/#fact-cyberhaven-20) | [None found](/vendors/cyberhaven/#fact-cyberhaven-20) | [Yes](/vendors/cyberhaven/#fact-cyberhaven-17) | [Partial](/vendors/cyberhaven/#fact-cyberhaven-19) | | [Grip Security](/vendors/grip-security/) | [Partial](/vendors/grip-security/#fact-grip-security-18) | [Partial](/vendors/grip-security/#fact-grip-security-19) | [Partial](/vendors/grip-security/#fact-grip-security-18) | [Partial](/vendors/grip-security/#fact-grip-security-17) | [None found](/vendors/grip-security/#fact-grip-security-20) | [Partial](/vendors/grip-security/#fact-grip-security-19) | | [Harmonic Security](/vendors/harmonic-security/) | [Yes](/vendors/harmonic-security/#fact-harmonic-security-28) | [Yes](/vendors/harmonic-security/#fact-harmonic-security-28) | [Partial](/vendors/harmonic-security/#fact-harmonic-security-32) | None found | None found | [Partial](/vendors/harmonic-security/#fact-harmonic-security-30) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | [Yes](/vendors/layerx/#fact-layerx-20) | [Yes](/vendors/layerx/#fact-layerx-20) | [None found](/vendors/layerx/#fact-layerx-23) | [None found](/vendors/layerx/#fact-layerx-23) | [Yes](/vendors/layerx/#fact-layerx-20) | None found | | [Microsoft](/vendors/microsoft/) | None found | [Partial](/vendors/microsoft/#fact-microsoft-04) | None found | [Partial](/vendors/microsoft/#fact-microsoft-06) | [Partial](/vendors/microsoft/#fact-microsoft-09) | None found | | [Netskope](/vendors/netskope/) | [Yes](/vendors/netskope/#fact-netskope-17) | [Yes](/vendors/netskope/#fact-netskope-12) | None found | [None found](/vendors/netskope/#fact-netskope-22) | [Yes](/vendors/netskope/#fact-netskope-18) | [None found](/vendors/netskope/#fact-netskope-22) | | [Nudge Security](/vendors/nudge-security/) | [Yes](/vendors/nudge-security/#fact-nudge-security-15) | None found | [Yes](/vendors/nudge-security/#fact-nudge-security-18) | [Partial](/vendors/nudge-security/#fact-nudge-security-16) | [Yes](/vendors/nudge-security/#fact-nudge-security-17) | None found | | [Palo Alto Networks](/vendors/palo-alto-networks/) | [Yes](/vendors/palo-alto-networks/#fact-palo-alto-networks-19) | [Yes](/vendors/palo-alto-networks/#fact-palo-alto-networks-20) | None found | [Partial](/vendors/palo-alto-networks/#fact-palo-alto-networks-18) | [Yes](/vendors/palo-alto-networks/#fact-palo-alto-networks-20) | None found | | [Push Security](/vendors/push-security/) | [Yes](/vendors/push-security/#fact-push-security-18) | [Yes](/vendors/push-security/#fact-push-security-07) | None found | [None found](/vendors/push-security/#fact-push-security-21) | [Yes](/vendors/push-security/#fact-push-security-19) | [Partial](/vendors/push-security/#fact-push-security-20) | | [Reco](/vendors/reco/) | None found | [Partial](/vendors/reco/#fact-reco-10) | None found | None found | None found | [Partial](/vendors/reco/#fact-reco-11) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | [Yes](/vendors/sentinelone/#fact-sentinelone-20) | [Yes](/vendors/sentinelone/#fact-sentinelone-21) | None found | [None found](/vendors/sentinelone/#fact-sentinelone-24) | [Yes](/vendors/sentinelone/#fact-sentinelone-19) | [None found](/vendors/sentinelone/#fact-sentinelone-24) | | [WitnessAI](/vendors/witnessai/) | None found | [Yes](/vendors/witnessai/#fact-witnessai-14) | [Partial](/vendors/witnessai/#fact-witnessai-14) | None found | [Yes](/vendors/witnessai/#fact-witnessai-12) | None found | | [Zscaler](/vendors/zscaler/) | [Yes](/vendors/zscaler/#fact-zscaler-20) | [Yes](/vendors/zscaler/#fact-zscaler-05) | None found | [None found](/vendors/zscaler/#fact-zscaler-27) | [Yes](/vendors/zscaler/#fact-zscaler-05) | [None found](/vendors/zscaler/#fact-zscaler-27) | - **Yes**: documented specifically - **Partial**: documented in general terms, for some tools only, or as an alert without the action - **None found**: none found on 16 Nov 2026 ## Why the steps after discovery matter A finding that nobody acts on is only a list. Some vendors stop at an alert, some coach the user, and some block or redact. The ladder separates those so a buyer can match the tool to the response the team is able to run. Read each column on its own. No vendor documents revoking OAuth grants or routing to a ticket specifically for AI tools; those two columns hold only partial and none-found cells, so a team that needs either step should ask for a demonstration. ## Download and cite [Download the Finding-to-Action Ladder as CSV or JSON](/datasets/finding-to-action-ladder/). Cite it as: Shadow AI Detection Tools, Vantage Point Matrix and Finding-to-Action Ladder, edition of 16 November 2026, https://shadowaidetectiontools.com/, licensed CC BY 4.0. ## Related pages - [How the Finding-to-Action Ladder is built](/finding-to-action-ladder/methodology/) - [Finding-to-Action Ladder data download](/datasets/finding-to-action-ladder/) - [Warn or coach](/finding-to-action-ladder/warn-or-coach/) - [Block](/finding-to-action-ladder/block/) - [Request approval](/finding-to-action-ladder/request-approval/) - [Revoke OAuth grant](/finding-to-action-ladder/revoke-oauth-grant/) - [Redact or stop data](/finding-to-action-ladder/redact-or-stop-data/) - [Route to ticket](/finding-to-action-ladder/route-to-ticket/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Vendor profiles](/vendors/) - [The ranking](/ranking/) --- # Finding-to-Action Ladder methodology URL: https://shadowaidetectiontools.com/finding-to-action-ladder/methodology/ Updated: 2026-11-16 The Finding-to-Action Ladder records, for 16 shadow AI detection vendors, which of six actions each documents for a discovered AI tool: warn or coach, block, require approval, revoke OAuth grants, redact or stop data, route to a ticket. One written rule sets every cell, and each cell cites fact ids from pages read on 16 Nov 2026. ## What the ladder is Finding an AI tool is the first step. The ladder lists what a vendor documents doing next, in six steps, so a buyer can see how far each tool goes from a finding to a response. ## The rule Each cell is derived from the facts tagged for the actions criterion in that vendor's evidence file. Yes means a fact names that action specifically for discovered AI tools. Partial means the action is named in general terms, for some tools only, or as an alert without the action. None found means no fact names it. The rule is applied the same way to all 16 vendors, Bay included. Facts not tagged for the actions criterion do not raise a level. A vendor's claim is recorded as the vendor's claim, not as a tested result, and third-party reports are marked as such in the cell note. ## Sources Cells cite fact ids from the vendor evidence files: product pages, documentation, release notes and press releases, with the URL and read date held on each fact. Where a vendor's page was searched for an action and nothing was found, the cell cites the fact that records the search; where no such fact exists, the cell cites no fact. A cell-by-cell audit of the 96 cells, with the deciding words of each fact, is kept with the site content. ## Read date All pages were read on 16 Nov 2026. Later changes are not reflected until the ladder is next updated. ## How to cite Cite as: Shadow AI Detection Tools, Finding-to-Action Ladder, edition of 16 Nov 2026, https://shadowaidetectiontools.com/finding-to-action-ladder/, CC BY 4.0. ## License The ladder and its CSV and JSON files are published under CC BY 4.0. Reuse with the citation above. ## Corrections To correct a cell, send the vendor, the action and a public page that documents it. Accepted changes are logged on the corrections page with their date. ## Levels and counts - **Yes** (36 cells): documented specifically - **Partial** (20 cells): documented in general terms, for some tools only, or as an alert without the action - **None found** (40 cells): none found on 16 Nov 2026 ## Related pages - [The Finding-to-Action Ladder](/finding-to-action-ladder/) - [Finding-to-Action Ladder download](/datasets/finding-to-action-ladder/) - [Scoring method](/method/) - [Warn or coach](/finding-to-action-ladder/warn-or-coach/) - [Block](/finding-to-action-ladder/block/) - [Request approval](/finding-to-action-ladder/request-approval/) --- # Warn or coach users about shadow AI URL: https://shadowaidetectiontools.com/finding-to-action-ladder/warn-or-coach/ Updated: 2026-11-16 Warning or coaching tells the user, in the moment, that the tool is not approved and where to go instead. Across the 16 vendors read on 16 Nov 2026, 11 document it specifically, 2 only in part (in general terms, for some tools only or as an alert), and 3 show none found. Cells cite fact ids. ## Warn or coach: every vendor **Warn or coach by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aurascape](/vendors/aurascape/) | Yes | Aurascape says it can coach users toward the enterprise tenant and coach with warning or confirmation notifications on restricted models. | [aurascape-23](/vendors/aurascape/#fact-aurascape-23), [aurascape-25](/vendors/aurascape/#fact-aurascape-25) | | [Bay](/vendors/bay/) | Yes | Bay says its browser guardrails let a team allow, warn, steer or block AI prompts. | [bay-23](/vendors/bay/#fact-bay-23) | | [Cyberhaven](/vendors/cyberhaven/) | Yes | Cyberhaven says it coaches employees with plain-English policy explanations and shows a policy message with approved alternatives for unapproved AI apps. | [cyberhaven-16](/vendors/cyberhaven/#fact-cyberhaven-16), [cyberhaven-18](/vendors/cyberhaven/#fact-cyberhaven-18) | | [Harmonic Security](/vendors/harmonic-security/) | Yes | Harmonic says Guide gives inline coaching that warns or nudges and adds warning banners for AI app tiers. | [harmonic-security-28](/vendors/harmonic-security/#fact-harmonic-security-28), [harmonic-security-31](/vendors/harmonic-security/#fact-harmonic-security-31) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Yes | Akamai says AI usage policies can warn or guide user actions based on context and risk. | [layerx-20](/vendors/layerx/#fact-layerx-20) | | [Netskope](/vendors/netskope/) | Yes | Netskope says real-time user coaching explains the risk and points to the approved option. | [netskope-17](/vendors/netskope/#fact-netskope-17), [netskope-20](/vendors/netskope/#fact-netskope-20) | | [Nudge Security](/vendors/nudge-security/) | Yes | Nudge Security says its browser extension nudges employees from unapproved apps to sanctioned tools and delivers the AI acceptable use policy. | [nudge-security-15](/vendors/nudge-security/#fact-nudge-security-15) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Yes | Palo Alto Networks says AI Access Security shows notifications and coaching when employees try unsanctioned GenAI apps. | [palo-alto-networks-19](/vendors/palo-alto-networks/#fact-palo-alto-networks-19) | | [Push Security](/vendors/push-security/) | Yes | Push says acknowledge mode warns users and points them to approved alternatives instead of blocking. | [push-security-18](/vendors/push-security/#fact-push-security-18) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Yes | SentinelOne says AI Usage Control coaches employees in the moment instead of blocking their work. | [sentinelone-20](/vendors/sentinelone/#fact-sentinelone-20) | | [Zscaler](/vendors/zscaler/) | Yes | Zscaler says it pairs with Workflow Automation to coach users on AI usage best practices. | [zscaler-20](/vendors/zscaler/#fact-zscaler-20) | | [Aim Security](/vendors/aim-security/) | Partial | Cato describes a pop-up telling the user when a prompt to an AI service is stopped and lists redirecting users to approved tools; a warn or coach mode is not named. | [aim-security-23](/vendors/aim-security/#fact-aim-security-23), [aim-security-25](/vendors/aim-security/#fact-aim-security-25) | | [Grip Security](/vendors/grip-security/) | Partial | Grip lists notifying on unsanctioned app signups and targeted user prompts for apps in general, not AI tools specifically. | [grip-security-18](/vendors/grip-security/#fact-grip-security-18) | | [Microsoft](/vendors/microsoft/) | None found | None found on 16 Nov 2026 in the actions facts for a user warning or coaching on AI tools. | none | | [Reco](/vendors/reco/) | None found | None found on 16 Nov 2026 in the actions facts for a user warning or coaching. | none | | [WitnessAI](/vendors/witnessai/) | None found | None found on 16 Nov 2026 in the actions facts for a user warning or coaching. | none | ## What warn or coach means here Warning or coaching tells the user, in the moment, that the tool is not approved and where to go instead. Coaching keeps the employee working and teaches the policy at the point of use. Netskope says real-time coaching explains the risk and points to the approved option. SentinelOne says AI Usage Control coaches employees in the moment instead of blocking their work. Push says acknowledge mode warns users and points them to approved alternatives. Sources: [Netskope source (netskope-17)](https://www.netskope.com/products/ai-products); [SentinelOne (Prompt Security) source (sentinelone-20)](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/); [Push Security source (push-security-18)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## Where the documented action stops short Warn and coach actions depend on a surface that can show a message: a browser extension, an endpoint client or an inline proxy page. A tool that reads only identity or API data can alert an administrator but has no screen of its own to coach on, which is why those vendors show none found or partial. Sources: [Push Security source (push-security-18)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## Vendors by level Yes (11): Aurascape, Bay, Cyberhaven, Harmonic Security, LayerX, Netskope, Nudge Security, Palo Alto Networks, Push Security, SentinelOne and Zscaler. Partial (2): Aim Security and Grip Security. None found on 16 Nov 2026 (3): Microsoft, Reco and WitnessAI. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Route to ticket](/finding-to-action-ladder/route-to-ticket/) - [Block](/finding-to-action-ladder/block/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Block discovered AI tools: which tools can URL: https://shadowaidetectiontools.com/finding-to-action-ladder/block/ Updated: 2026-11-16 Blocking stops the use of a discovered AI tool, a domain, an extension or a risky action inside the tool. Across the 16 vendors read on 16 Nov 2026, 12 document it specifically, 3 only in part (in general terms, for some tools only or as an alert), and 1 shows none found. Cells cite fact ids. ## Block: every vendor **Block by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | Yes | Cato says browser and desktop AI rules can block a prompt, and coding agents get block controls. | [aim-security-24](/vendors/aim-security/#fact-aim-security-24), [aim-security-42](/vendors/aim-security/#fact-aim-security-42) | | [Aurascape](/vendors/aurascape/) | Yes | Aurascape says policy can block a risky AI tool entirely or block use of a restricted model. | [aurascape-21](/vendors/aurascape/#fact-aurascape-21), [aurascape-25](/vendors/aurascape/#fact-aurascape-25) | | [Bay](/vendors/bay/) | Yes | Bay says it can block specific AI domains and extensions and stop risky agentic actions in the browser. | [bay-24](/vendors/bay/#fact-bay-24), [bay-23](/vendors/bay/#fact-bay-23) | | [Cyberhaven](/vendors/cyberhaven/) | Yes | Cyberhaven says it blocks high-risk data movement at the AI prompt and response level and offers block controls for agent actions. | [cyberhaven-16](/vendors/cyberhaven/#fact-cyberhaven-16), [cyberhaven-17](/vendors/cyberhaven/#fact-cyberhaven-17) | | [Harmonic Security](/vendors/harmonic-security/) | Yes | Harmonic says Guide blocks inline, and its MCP Gateway blocks an attempt to use an MCP server the user is not permitted to use. | [harmonic-security-28](/vendors/harmonic-security/#fact-harmonic-security-28), [harmonic-security-32](/vendors/harmonic-security/#fact-harmonic-security-32) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Yes | Akamai says AI usage policies can block user actions and it blocks malicious or overprivileged browser extensions. | [layerx-20](/vendors/layerx/#fact-layerx-20), [layerx-22](/vendors/layerx/#fact-layerx-22) | | [Netskope](/vendors/netskope/) | Yes | Netskope says it blocks threats in generative AI apps and can default-block public MCP servers. | [netskope-12](/vendors/netskope/#fact-netskope-12), [netskope-19](/vendors/netskope/#fact-netskope-19) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Yes | Palo Alto Networks says AI Access Security inline detection blocks sensitive text and file transfers to GenAI apps. | [palo-alto-networks-20](/vendors/palo-alto-networks/#fact-palo-alto-networks-20) | | [Push Security](/vendors/push-security/) | Yes | Push says custom detections can block unapproved MCP server connections in the browser and block mode stops logins to unapproved apps. | [push-security-07](/vendors/push-security/#fact-push-security-07), [push-security-17](/vendors/push-security/#fact-push-security-17) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Yes | Techzine, a third-party report, says Prompt Security policy controls block risky prompts before they are sent. | [sentinelone-21](/vendors/sentinelone/#fact-sentinelone-21) | | [WitnessAI](/vendors/witnessai/) | Yes | WitnessAI says a tool or MCP server not on the approved list is denied before the tool executes. | [witnessai-14](/vendors/witnessai/#fact-witnessai-14) | | [Zscaler](/vendors/zscaler/) | Yes | Zscaler says it blocks AI access by role in real time and enforces which AI tools users can access. | [zscaler-05](/vendors/zscaler/#fact-zscaler-05), [zscaler-18](/vendors/zscaler/#fact-zscaler-18) | | [Grip Security](/vendors/grip-security/) | Partial | Grip lists a Cisco Umbrella integration that blocks risky or unapproved apps it discovers, apps in general, and names no in-the-moment blocking of AI prompts. | [grip-security-19](/vendors/grip-security/#fact-grip-security-19), [grip-security-20](/vendors/grip-security/#fact-grip-security-20) | | [Microsoft](/vendors/microsoft/) | Partial | Microsoft documents blocking an unsanctioned discovered app through a proxy or firewall and blocking uploads from unallowed browsers, for apps in general. | [microsoft-04](/vendors/microsoft/#fact-microsoft-04), [microsoft-67](/vendors/microsoft/#fact-microsoft-67) | | [Reco](/vendors/reco/) | Partial | The Hacker News review, a third-party report, says Reco cannot prevent shadow AI tools from connecting but can identify and alert on the connections. | [reco-10](/vendors/reco/#fact-reco-10) | | [Nudge Security](/vendors/nudge-security/) | None found | None found on 16 Nov 2026 in the actions facts for blocking AI tools. | none | ## What block means here Blocking stops the use of a discovered AI tool, a domain, an extension or a risky action inside the tool. Several vendors document blocking at different depths. Aurascape says policy can block a risky AI tool entirely or only unsafe actions inside less risky tools. Push says block mode stops logins to unapproved apps with a blocking banner. WitnessAI says a tool or MCP server not on the approved list is denied before it executes. Sources: [Aurascape source (aurascape-21)](https://aurascape.ai/frictionless-ai-security/); [Push Security source (push-security-17)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [WitnessAI source (witnessai-14)](https://witness.ai/blog/introducing-witnessai-agentic-control-one-control-plane-for-every-agent-tool-and-mcp-server/). ## Where the documented action stops short Blocking by a visibility-first vendor is often indirect or general. Grip lists a Cisco Umbrella integration that blocks apps Grip discovers, and Microsoft blocks unsanctioned apps through a proxy or firewall; both cover apps in general, so both cells are partial. A review says Reco cannot prevent third-party AI tools from connecting but alerts on them, which is also partial. No Nudge Security actions fact names a block, so that cell shows none found. Sources: [Grip Security source (grip-security-19)](https://www.grip.security/integrations); [Microsoft source (microsoft-04)](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Reco source (reco-10)](https://thehackernews.com/2025/01/product-review-how-reco-discovers.html). ## Vendors by level Yes (12): Aim Security, Aurascape, Bay, Cyberhaven, Harmonic Security, LayerX, Netskope, Palo Alto Networks, Push Security, SentinelOne, WitnessAI and Zscaler. Partial (3): Grip Security, Microsoft and Reco. None found on 16 Nov 2026 (1): Nudge Security. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Warn or coach](/finding-to-action-ladder/warn-or-coach/) - [Request approval](/finding-to-action-ladder/request-approval/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Request approval for newly found AI tools URL: https://shadowaidetectiontools.com/finding-to-action-ladder/request-approval/ Updated: 2026-11-16 Request approval means a newly found AI tool enters a state that needs an administrator's decision before use. Across the 16 vendors read on 16 Nov 2026, 1 documents it specifically, 5 only in part (in general terms, for some tools only or as an alert), and 10 show none found. Cells cite fact ids. ## Request approval: every vendor **Request approval by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Nudge Security](/vendors/nudge-security/) | Yes | Nudge Security lets admins set an approval status of Approved, Allowed, In Review or Not Permitted for each discovered agent. | [nudge-security-18](/vendors/nudge-security/#fact-nudge-security-18) | | [Aurascape](/vendors/aurascape/) | Partial | Aurascape says users can request limited-time exceptions to a block; an approval step for newly found tools is not named. | [aurascape-22](/vendors/aurascape/#fact-aurascape-22) | | [Bay](/vendors/bay/) | Partial | Bay names an approval workflow for newly found AI tools only as a page heading; none found on 16 Nov 2026 for how the approval step works, so the action is named in general terms. | [bay-25](/vendors/bay/#fact-bay-25) | | [Grip Security](/vendors/grip-security/) | Partial | Grip lists approval workflows for app onboarding, for apps in general, not AI tools specifically. | [grip-security-18](/vendors/grip-security/#fact-grip-security-18) | | [Harmonic Security](/vendors/harmonic-security/) | Partial | Harmonic's docs say a blocked attempt to use an MCP server creates an access request for an administrator, for MCP servers only. | [harmonic-security-32](/vendors/harmonic-security/#fact-harmonic-security-32) | | [WitnessAI](/vendors/witnessai/) | Partial | WitnessAI names an approved list for agent tools and MCP servers only; an approval step for discovered AI apps is not named. | [witnessai-14](/vendors/witnessai/#fact-witnessai-14) | | [Aim Security](/vendors/aim-security/) | None found | None found on 16 Nov 2026 in the actions facts for an approval step for AI tools. | none | | [Cyberhaven](/vendors/cyberhaven/) | None found | None found on 16 Nov 2026 for an approval workflow before an AI tool is used. | [cyberhaven-20](/vendors/cyberhaven/#fact-cyberhaven-20) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | None found | None found on 16 Nov 2026 for an approval workflow for AI tools. | [layerx-23](/vendors/layerx/#fact-layerx-23) | | [Microsoft](/vendors/microsoft/) | None found | None found on 16 Nov 2026 in the actions facts for an approval step for AI tools. | none | | [Netskope](/vendors/netskope/) | None found | None found on 16 Nov 2026 in the actions facts for an approval workflow for AI tools. | none | | [Palo Alto Networks](/vendors/palo-alto-networks/) | None found | None found on 16 Nov 2026 in the actions facts for an approval workflow. | none | | [Push Security](/vendors/push-security/) | None found | None found on 16 Nov 2026 in the actions facts for an approval request flow. | none | | [Reco](/vendors/reco/) | None found | None found on 16 Nov 2026 in the actions facts for an approval step. | none | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | None found | None found on 16 Nov 2026 in the actions facts for an approval flow. | none | | [Zscaler](/vendors/zscaler/) | None found | None found on 16 Nov 2026 in the actions facts for an approval request flow. | none | ## What request approval means here Request approval means a newly found AI tool enters a state that needs an administrator's decision before use. Nudge Security lets admins set Approved, Allowed, In Review or Not Permitted for each discovered agent. Bay names an approval workflow for newly found AI tools only as a heading on its page, with no detail of how the step runs, so its cell is partial. Grip lists approval workflows for app onboarding in general, so its cell is partial. Sources: [Nudge Security source (nudge-security-18)](https://www.nudgesecurity.com/features/ai-agent-discovery); [Grip Security source (grip-security-18)](https://www.grip.security/ai-security); [Bay source (bay-25)](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=asset-dim). ## Where the documented action stops short Several vendors name approval for some tools only. Harmonic's docs say a blocked attempt to use an MCP server creates an access request, WitnessAI's approved list covers agent tools and MCP servers, and Aurascape lets users request limited-time exceptions to a block. All three are recorded as partial because an approval step for a newly found AI tool is not named. Sources: [Harmonic Security source (harmonic-security-32)](https://docs.harmonicsecurity.app/portal-guides/managing-your-organization/configure-mcp-server-access-controls); [WitnessAI source (witnessai-14)](https://witness.ai/blog/introducing-witnessai-agentic-control-one-control-plane-for-every-agent-tool-and-mcp-server/); [Aurascape source (aurascape-22)](https://aurascape.ai/frictionless-ai-security/). ## Vendors by level Yes (1): Nudge Security. Partial (5): Aurascape, Bay, Grip Security, Harmonic Security and WitnessAI. None found on 16 Nov 2026 (10): Aim Security, Cyberhaven, LayerX, Microsoft, Netskope, Palo Alto Networks, Push Security, Reco, SentinelOne and Zscaler. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Block](/finding-to-action-ladder/block/) - [Revoke OAuth grant](/finding-to-action-ladder/revoke-oauth-grant/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Revoke OAuth grants held by AI apps URL: https://shadowaidetectiontools.com/finding-to-action-ladder/revoke-oauth-grant/ Updated: 2026-11-16 Revoking an OAuth grant removes the access an AI app was given to a work account, such as Google Workspace or Microsoft 365. Across the 16 vendors read on 16 Nov 2026, none documents it specifically, 4 only in part (in general terms, for some tools only or as an alert), and 12 show none found. Cells cite fact ids. ## Revoke OAuth grant: every vendor **Revoke OAuth grant by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Grip Security](/vendors/grip-security/) | Partial | Grip says it can revoke risky access through automated password rotation for apps in general; OAuth grants of AI tools are not named. | [grip-security-17](/vendors/grip-security/#fact-grip-security-17) | | [Microsoft](/vendors/microsoft/) | Partial | Microsoft says app governance shows the permissions OAuth apps hold and supports blocking or restricting risky apps, apps in general; revoking grants is not named. | [microsoft-06](/vendors/microsoft/#fact-microsoft-06) | | [Nudge Security](/vendors/nudge-security/) | Partial | Nudge Security lists auto-revoking overly permissive OAuth grants as a paid-tier item for third-party grants in general, not AI apps specifically. | [nudge-security-16](/vendors/nudge-security/#fact-nudge-security-16) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Partial | Palo Alto Networks says administrators can revoke access by privilege scope and risk; OAuth grants are not named. | [palo-alto-networks-18](/vendors/palo-alto-networks/#fact-palo-alto-networks-18) | | [Aim Security](/vendors/aim-security/) | None found | None found on 16 Nov 2026 in the actions facts for revoking OAuth grants. | none | | [Aurascape](/vendors/aurascape/) | None found | None found on 16 Nov 2026 for revoking OAuth grants. | [aurascape-26](/vendors/aurascape/#fact-aurascape-26) | | [Bay](/vendors/bay/) | None found | None found on 16 Nov 2026 for revoking OAuth grants. | [bay-32](/vendors/bay/#fact-bay-32) | | [Cyberhaven](/vendors/cyberhaven/) | None found | None found on 16 Nov 2026 for revoking OAuth grants to AI apps. | [cyberhaven-20](/vendors/cyberhaven/#fact-cyberhaven-20) | | [Harmonic Security](/vendors/harmonic-security/) | None found | None found on 16 Nov 2026 in the actions facts for revoking OAuth grants. | none | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | None found | None found on 16 Nov 2026 for OAuth grant revocation. | [layerx-23](/vendors/layerx/#fact-layerx-23) | | [Netskope](/vendors/netskope/) | None found | None found on 16 Nov 2026 for revoking OAuth grants. | [netskope-22](/vendors/netskope/#fact-netskope-22) | | [Push Security](/vendors/push-security/) | None found | None found on 16 Nov 2026 for revoking an OAuth grant held by an AI app. | [push-security-21](/vendors/push-security/#fact-push-security-21) | | [Reco](/vendors/reco/) | None found | None found on 16 Nov 2026 in the actions facts for revoking OAuth grants. | none | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | None found | None found on 16 Nov 2026 for OAuth grant revocation. | [sentinelone-24](/vendors/sentinelone/#fact-sentinelone-24) | | [WitnessAI](/vendors/witnessai/) | None found | None found on 16 Nov 2026 in the actions facts for revoking OAuth grants. | none | | [Zscaler](/vendors/zscaler/) | None found | None found on 16 Nov 2026 for revoking OAuth grants. | [zscaler-27](/vendors/zscaler/#fact-zscaler-27) | ## What revoke oauth grant means here Revoking an OAuth grant removes the access an AI app was given to a work account, such as Google Workspace or Microsoft 365. This is the thinnest step in the ladder, and no vendor documents it specifically for AI apps. Grip Security says it can revoke risky access through automated password rotation, for apps in general. Nudge Security lists auto-revoking overly permissive OAuth grants as a paid-tier item, though its facts do not limit it to AI apps. Palo Alto Networks says AI Access Security administrators can revoke access by privilege scope and risk, without naming OAuth grants. Microsoft's app governance shows the permissions OAuth apps hold and supports blocking or restricting risky apps. Sources: [Grip Security source (grip-security-17)](https://www.grip.security/use-case-library/reduce-shadow-ai-risks); [Nudge Security source (nudge-security-16)](https://www.nudgesecurity.com/pricing); [Palo Alto Networks source (palo-alto-networks-18)](https://www.paloaltonetworks.com/sase/ai-access-security); [Microsoft source (microsoft-06)](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance). ## Where the documented action stops short Eight vendors have a fact recording that revoking OAuth grants was looked for on their pages and not found on 16 Nov 2026: Aurascape, Bay, Cyberhaven, LayerX, Netskope, Push Security, SentinelOne and Zscaler. For the other four none-found cells, no actions fact names the step. Sources: [Push Security source (push-security-21)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [Bay source (bay-32)](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=asset-dim); [Netskope source (netskope-22)](https://www.netskope.com/products/ai-command-center). ## Vendors by level Yes (0): none. Partial (4): Grip Security, Microsoft, Nudge Security and Palo Alto Networks. None found on 16 Nov 2026 (12): Aim Security, Aurascape, Bay, Cyberhaven, Harmonic Security, LayerX, Netskope, Push Security, Reco, SentinelOne, WitnessAI and Zscaler. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Request approval](/finding-to-action-ladder/request-approval/) - [Redact or stop data](/finding-to-action-ladder/redact-or-stop-data/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Redact or stop data sent to AI tools URL: https://shadowaidetectiontools.com/finding-to-action-ladder/redact-or-stop-data/ Updated: 2026-11-16 Redact or stop data means sensitive text or files are removed, masked or held back before they reach the AI tool. Across the 16 vendors read on 16 Nov 2026, 12 document it specifically, 1 only in part (in general terms, for some tools only or as an alert), and 3 show none found. Cells cite fact ids. ## Redact or stop data: every vendor **Redact or stop data by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Aim Security](/vendors/aim-security/) | Yes | Cato says rules can block or redact sensitive data and anonymize the sensitive data in a browser prompt. | [aim-security-22](/vendors/aim-security/#fact-aim-security-22), [aim-security-42](/vendors/aim-security/#fact-aim-security-42) | | [Aurascape](/vendors/aurascape/) | Yes | Aurascape says it can redact sensitive data in a request and, from its September 2026 release, mask values before they reach the model. | [aurascape-23](/vendors/aurascape/#fact-aurascape-23), [aurascape-24](/vendors/aurascape/#fact-aurascape-24) | | [Bay](/vendors/bay/) | Yes | Bay says its browser guardrails keep personal data, secrets and source code out of prompts. | [bay-23](/vendors/bay/#fact-bay-23) | | [Cyberhaven](/vendors/cyberhaven/) | Yes | Cyberhaven says it offers block, warn and redact controls for risky agent actions across the agentic workflow. | [cyberhaven-17](/vendors/cyberhaven/#fact-cyberhaven-17) | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | Yes | Akamai says AI usage policies can redact and that enforcement happens before data leaves the user. | [layerx-20](/vendors/layerx/#fact-layerx-20), [layerx-21](/vendors/layerx/#fact-layerx-21) | | [Netskope](/vendors/netskope/) | Yes | Netskope says redaction and blocking stop sensitive data before it reaches a model. | [netskope-18](/vendors/netskope/#fact-netskope-18), [netskope-21](/vendors/netskope/#fact-netskope-21) | | [Nudge Security](/vendors/nudge-security/) | Yes | Nudge Security says employees can redact detected sensitive data in one click and admins can enforce redaction in the browser (9 Nov 2026). | [nudge-security-17](/vendors/nudge-security/#fact-nudge-security-17) | | [Palo Alto Networks](/vendors/palo-alto-networks/) | Yes | Palo Alto Networks says inline detection blocks sensitive text and file transfers to GenAI apps. | [palo-alto-networks-20](/vendors/palo-alto-networks/#fact-palo-alto-networks-20) | | [Push Security](/vendors/push-security/) | Yes | Push says clipboard warn mode can offer a redacted version of sensitive data, and its June 2026 launch added clipboard controls for unsanctioned AI tools. | [push-security-19](/vendors/push-security/#fact-push-security-19), [push-security-35](/vendors/push-security/#fact-push-security-35) | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | Yes | SentinelOne says AI Usage Control redacts PII, secrets and other sensitive data inline before a prompt leaves the device. | [sentinelone-19](/vendors/sentinelone/#fact-sentinelone-19) | | [WitnessAI](/vendors/witnessai/) | Yes | WitnessAI says it redacts sensitive data in real time and tokenizes PII, credentials and secrets before they reach a model. | [witnessai-12](/vendors/witnessai/#fact-witnessai-12), [witnessai-13](/vendors/witnessai/#fact-witnessai-13) | | [Zscaler](/vendors/zscaler/) | Yes | Zscaler says it filters sensitive data sent to LLMs and can disable cut, paste and download to GenAI apps in an isolated browser. | [zscaler-05](/vendors/zscaler/#fact-zscaler-05), [zscaler-19](/vendors/zscaler/#fact-zscaler-19) | | [Microsoft](/vendors/microsoft/) | Partial | Microsoft documents detection of sensitive information pasted or shared with AI apps; an action that redacts or stops it is not named. | [microsoft-09](/vendors/microsoft/#fact-microsoft-09), [microsoft-10](/vendors/microsoft/#fact-microsoft-10) | | [Grip Security](/vendors/grip-security/) | None found | None found on 16 Nov 2026: Grip's pages name no redaction of data sent to AI tools. | [grip-security-20](/vendors/grip-security/#fact-grip-security-20) | | [Harmonic Security](/vendors/harmonic-security/) | None found | None found on 16 Nov 2026 in the actions facts for redacting or stopping sensitive data sent to an AI tool. | none | | [Reco](/vendors/reco/) | None found | None found on 16 Nov 2026 in the actions facts for redacting or stopping data sent to a tool. | none | ## What redact or stop data means here Redact or stop data means sensitive text or files are removed, masked or held back before they reach the AI tool. SentinelOne says AI Usage Control redacts PII, secrets and other sensitive data inline on the device before a prompt leaves it. Aurascape's September 2026 release adds masking of sensitive values in prompts, which restores the real value in the reply. Nudge Security says employees can redact detected sensitive data in one click and admins can enforce redaction (9 Nov 2026). Sources: [SentinelOne (Prompt Security) source (sentinelone-19)](https://www.sentinelone.com/platform/prompt-security-ai-usage-control/); [Aurascape source (aurascape-24)](https://aurascape.ai/resources/blog/september-2026-product-update/); [Nudge Security source (nudge-security-17)](https://www.nudgesecurity.com/changelog). ## Where the documented action stops short Detection alone is recorded as partial. Microsoft documents detection of sensitive information shared with AI apps and of pasting to restricted domains; an action that redacts it is not named in those facts. Grip, Harmonic Security and Reco show none found because no actions fact names redacting or stopping data. Sources: [Microsoft source (microsoft-09)](https://learn.microsoft.com/en-us/purview/dspm-for-ai); [Microsoft source (microsoft-10)](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about); [Grip Security source (grip-security-20)](https://www.grip.security/use-case-library/prevent-sensitive-data-exposure-to-ai). ## Vendors by level Yes (12): Aim Security, Aurascape, Bay, Cyberhaven, LayerX, Netskope, Nudge Security, Palo Alto Networks, Push Security, SentinelOne, WitnessAI and Zscaler. Partial (1): Microsoft. None found on 16 Nov 2026 (3): Grip Security, Harmonic Security and Reco. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Revoke OAuth grant](/finding-to-action-ladder/revoke-oauth-grant/) - [Route to ticket](/finding-to-action-ladder/route-to-ticket/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Route a shadow AI finding to a ticket URL: https://shadowaidetectiontools.com/finding-to-action-ladder/route-to-ticket/ Updated: 2026-11-16 Route to ticket means the finding becomes a task in the customer's ticketing or workflow system for an owner to resolve. Across the 16 vendors read on 16 Nov 2026, none documents it specifically, 5 only in part (in general terms, for some tools only or as an alert), and 11 show none found. Cells cite fact ids. ## Route to ticket: every vendor **Route to ticket by vendor** | Vendor | Level | Note | Evidence | |---|---|---|---| | [Cyberhaven](/vendors/cyberhaven/) | Partial | Cyberhaven says it streams incidents and events to SIEM, SOAR and ServiceNow integrations; creating a ticket for a discovered AI tool is not named. | [cyberhaven-19](/vendors/cyberhaven/#fact-cyberhaven-19) | | [Grip Security](/vendors/grip-security/) | Partial | Grip lists a Jira integration that creates tickets for shadow SaaS management in general, not AI tools specifically. | [grip-security-19](/vendors/grip-security/#fact-grip-security-19) | | [Harmonic Security](/vendors/harmonic-security/) | Partial | Harmonic's webhook docs list alert and intervention events that can feed workflows such as ticketing; ticket creation itself is not named. | [harmonic-security-30](/vendors/harmonic-security/#fact-harmonic-security-30) | | [Push Security](/vendors/push-security/) | Partial | Push documents a REST API and webhooks that can create tickets in Jira for events in general, not AI tools specifically. | [push-security-20](/vendors/push-security/#fact-push-security-20) | | [Reco](/vendors/reco/) | Partial | Reco says its SaaS ticketing workflow creates and tracks remediation tasks in the existing ticketing system, for SaaS issues in general. | [reco-11](/vendors/reco/#fact-reco-11) | | [Aim Security](/vendors/aim-security/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [Aurascape](/vendors/aurascape/) | None found | None found on 16 Nov 2026 for routing a discovered AI tool to a ticketing system. | [aurascape-26](/vendors/aurascape/#fact-aurascape-26) | | [Bay](/vendors/bay/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [LayerX (Akamai Workforce Protector)](/vendors/layerx/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [Microsoft](/vendors/microsoft/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [Netskope](/vendors/netskope/) | None found | None found on 16 Nov 2026 for routing a discovered AI tool to a ticket. | [netskope-22](/vendors/netskope/#fact-netskope-22) | | [Nudge Security](/vendors/nudge-security/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [Palo Alto Networks](/vendors/palo-alto-networks/) | None found | None found on 16 Nov 2026 in the actions facts for routing a finding to a ticket. | none | | [SentinelOne (Prompt Security)](/vendors/sentinelone/) | None found | None found on 16 Nov 2026 for routing to a ticket. | [sentinelone-24](/vendors/sentinelone/#fact-sentinelone-24) | | [WitnessAI](/vendors/witnessai/) | None found | None found on 16 Nov 2026 in the actions facts for routing to a ticket. | none | | [Zscaler](/vendors/zscaler/) | None found | None found on 16 Nov 2026 for routing a discovered AI app to a ticket. | [zscaler-27](/vendors/zscaler/#fact-zscaler-27) | ## What route to ticket means here Route to ticket means the finding becomes a task in the customer's ticketing or workflow system for an owner to resolve. Reco says its SaaS ticketing workflow creates and tracks remediation tasks in the customer's existing ticketing system, and Grip lists a Jira integration that creates tickets for shadow SaaS management. Both cover SaaS in general, not AI tools specifically, so both cells are partial. Push documents a REST API and webhooks that can create tickets in Jira or similar systems for events in general. Sources: [Reco source (reco-11)](https://www.reco.ai/use-cases/shadow-ai-discovery); [Grip Security source (grip-security-19)](https://www.grip.security/integrations); [Push Security source (push-security-20)](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system). ## Where the documented action stops short Streaming to a SIEM is not the same step. Cyberhaven streams incidents to SIEM, SOAR and AI SOC platforms with Splunk, Cortex XSOAR and ServiceNow integrations, and Harmonic's webhooks can feed ticketing; both are recorded as partial. Sources: [Push Security source (push-security-20)](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system); [Cyberhaven source (cyberhaven-19)](https://www.cyberhaven.com/product/integrations); [Harmonic Security source (harmonic-security-30)](https://docs.harmonicsecurity.app/portal-guides/integration-guides/configure-a-webhook). ## Vendors by level Yes (0): none. Partial (5): Cyberhaven, Grip Security, Harmonic Security, Push Security and Reco. None found on 16 Nov 2026 (11): Aim Security, Aurascape, Bay, LayerX, Microsoft, Netskope, Nudge Security, Palo Alto Networks, SentinelOne, WitnessAI and Zscaler. ## Related pages - [The full Finding-to-Action Ladder](/finding-to-action-ladder/) - [Redact or stop data](/finding-to-action-ladder/redact-or-stop-data/) - [Warn or coach](/finding-to-action-ladder/warn-or-coach/) - [Method](/finding-to-action-ladder/methodology/) - [Vendor profiles](/vendors/) --- # Shadow AI detection datasets URL: https://shadowaidetectiontools.com/datasets/ Updated: 2026-11-16 Two datasets are available as CSV and JSON under CC BY 4.0: the Vantage Point Matrix, which records which of six vantage points each of 16 vendors reads, and the Finding-to-Action Ladder, which records six actions each can take. Cells cite fact IDs read on 16 Nov 2026. ## Downloads - [Vantage Point Matrix](/datasets/vantage-point-matrix/): [CSV](/datasets/vantage-point-matrix.csv), [JSON](/datasets/vantage-point-matrix.json) - [Finding-to-Action Ladder](/datasets/finding-to-action-ladder/): [CSV](/datasets/finding-to-action-ladder.csv), [JSON](/datasets/finding-to-action-ladder.json) ## License and citation CC BY 4.0. Shadow AI Detection Tools, Vantage Point Matrix and Finding-to-Action Ladder, edition of 16 November 2026, https://shadowaidetectiontools.com/, licensed CC BY 4.0. ## Related pages - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) - [Method](/method/) - [The ranking](/ranking/) --- # Vantage Point Matrix dataset: CSV and JSON download URL: https://shadowaidetectiontools.com/datasets/vantage-point-matrix/ Updated: 2026-11-16 The Vantage Point Matrix is published as a CSV file and a JSON file under CC BY 4.0. It holds 80 vendor and dimension cells across 5 dimensions, read on 2026-11-16. Cite it with the line below when you reuse the data, and link back to this page so readers can check the files against the sources. ## Files - [Vantage Point Matrix as CSV](/datasets/vantage-point-matrix.csv) - [Vantage Point Matrix as JSON](/datasets/vantage-point-matrix.json) ## How to cite Shadow AI Detection Tools, Vantage Point Matrix and Finding-to-Action Ladder, edition of 16 November 2026, https://shadowaidetectiontools.com/, licensed CC BY 4.0. ## Related pages - [The Vantage Point Matrix](/vantage-point-matrix/) - [Method](/vantage-point-matrix/methodology/) - [All datasets](/datasets/) - [Scoring method](/method/) --- # Finding-to-Action Ladder dataset: CSV and JSON download URL: https://shadowaidetectiontools.com/datasets/finding-to-action-ladder/ Updated: 2026-11-16 The Finding-to-Action Ladder is published as a CSV file and a JSON file under CC BY 4.0. It holds 96 vendor and dimension cells across 6 dimensions, read on 2026-11-16. Cite it with the line below when you reuse the data, and link back to this page so readers can check the files against the sources. ## Files - [Finding-to-Action Ladder as CSV](/datasets/finding-to-action-ladder.csv) - [Finding-to-Action Ladder as JSON](/datasets/finding-to-action-ladder.json) ## How to cite Shadow AI Detection Tools, Vantage Point Matrix and Finding-to-Action Ladder, edition of 16 November 2026, https://shadowaidetectiontools.com/, licensed CC BY 4.0. ## Related pages - [The Finding-to-Action Ladder](/finding-to-action-ladder/) - [Method](/finding-to-action-ladder/methodology/) - [All datasets](/datasets/) - [Scoring method](/method/) --- # Shadow AI detection guides URL: https://shadowaidetectiontools.com/guides/ Updated: 2026-11-16 Twelve guides cover shadow AI detection in practice: running an audit, reading proxy, DNS and identity provider logs, finding AI embedded in approved SaaS, building and triaging the inventory, and reporting to auditors and the board. Each guide cites vendor documentation read on 16 Nov 2026. ## 12 guides - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/): How AI features appear inside approved SaaS apps and how to find them: admin consoles, feature flags, new OAuth grants and vendor sub-processor lists. - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/): A shadow AI audit in four steps: scope the kinds of AI, pull identity, network and endpoint data, reconcile it, and produce an inventory. - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/): How to tell a personal AI account from a work account: browser and identity signals, tenant restrictions, and what network logs cannot show. - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/): How to read proxy and DNS logs for AI use: which domains to look for, what the logs show per user, and what encryption and shared domains hide. - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/): What evidence auditors ask for on shadow AI: an inventory with dates, the sources behind it, decisions, change logs and retained detection records. - [Shadow AI Detection in Microsoft 365 and Google Workspace](/guides/shadow-ai-detection-in-microsoft-365-and-google-workspace/): What Microsoft Purview and Defender for Cloud Apps can see of shadow AI, what Google Workspace admin logs show, and where you need another source. - [Shadow AI Discovery From Identity Provider Logs](/guides/shadow-ai-discovery-from-identity-provider-logs/): How to find AI tools from identity provider logs: single sign-on events, OAuth consent records and app assignments in Entra ID, Google Workspace and Okta. - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/): What shadow AI detection tools get wrong: catalog gaps, shared domains, personal accounts and the kinds of AI use that stay invisible. - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/): Which shadow AI numbers belong in a board report: tools found, users, share on personal accounts, high-risk findings, time to resolve and coverage limits. - [Shadow AI vs Shadow IT: What Changes for Detection](/guides/shadow-ai-vs-shadow-it-what-changes-for-detection/): How shadow AI differs from shadow IT for detection: shared domains, embedded features, local agents and OAuth grants that classic SaaS discovery misses. - [Triaging What a Shadow AI Scan Finds](/guides/triaging-what-a-shadow-ai-scan-finds/): How to triage shadow AI scan results: rank findings by data exposure, account type, access scope and autonomy, then pick one action per finding. - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/): The fields a shadow AI inventory needs: tool, kind, owner, data, account type, source, risk, disposition and review date, with one owner per entry. ## Related pages - [Glossary](/glossary/) - [The ranking](/ranking/) - [FAQ](/faq/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # Embedded AI features switched on in approved SaaS URL: https://shadowaidetectiontools.com/guides/embedded-ai-features-switched-on-in-approved-saas/ Updated: 2026-11-16 An approved SaaS app can add an AI feature, an assistant or a new sub-processor without a new domain or install, so network and endpoint tools see nothing change. Find these features through SaaS admin consoles, OAuth and permission changes, and vendor change notices, or with a tool that reads SaaS APIs. ## How do AI features appear inside approved apps? They appear as a setting, a license add-on or an automatic update. The app, its domain and its sign-in stay the same, so nothing in a proxy log, DNS log or sign-in record changes. The data the app holds is now processed by a model, sometimes one run by a third party. Nudge Security says it shows which SaaS vendors use third-party AI services such as OpenAI and Anthropic, which it frames as AI in SaaS supply chains. Grip Security says it flags approved SaaS that has AI features. Sources: [Nudge Security pricing](https://www.nudgesecurity.com/pricing); [Grip Security: reduce shadow AI risks](https://www.grip.security/use-case-library/reduce-shadow-ai-risks). ## How do you find them? Check three places: the SaaS admin console for AI settings and add-ons, the change notices that vendors send before enabling a feature, and OAuth or API permission changes tied to the app. For the largest suites, review the AI section of the admin center each quarter and after each major release. Detection products read these records at scale. Zscaler documents a separate public AI inventory for embedded AI, Netskope says its AI Command Center covers embedded SaaS AI, and Harmonic Explore covers AI features embedded in SaaS tools. Reco and Grip Security list named embedded AI across their connected SaaS apps. Sources: [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai); [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center); [Harmonic Security Explore](https://www.harmonic.security/products/explore); [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery); [Grip Security: discover and control shadow AI](https://www.grip.security/use-case-library/discover-and-control-shadow-ai-across-your-enterprise); [Netskope source (netskope-06)](https://www.netskope.com/products/ai-command-center); [Reco source (reco-04)](https://www.reco.ai/platform/ai-governance-security). ## Which vantage points can see an embedded feature? SaaS APIs and the vendor's own admin and audit records can see an embedded feature, because the setting lives in the app. Identity logs see it only when the feature needs a new consent. Endpoint and network views see it only when the feature calls a new host. The Vantage Point Matrix on this site shows which products read SaaS APIs and logs. Microsoft documents SaaS posture sources in Defender for Cloud Apps, and Palo Alto Networks documents SaaS posture as one source for AI Access Security. Sources: [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance); [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps). ## What should the review decide? For each embedded feature decide whether it is enabled for everyone, a pilot group or no one, which data it can reach, and where inputs go. Ask the vendor whether inputs train a model and how long prompts are retained; Nudge Security says it summarizes each AI vendor's model-training policy, opt-out options and retention. Record the decision in the inventory against the app, not as a separate tool. An embedded feature is a property of an approved app and changes with it. Sources: [Nudge Security: AI discovery solution](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution). ## What controls help when a feature cannot be turned off? When a feature cannot be turned off, limit the data it can reach. Use app permissions and sensitivity labels, restrict which groups get the add-on, and watch the paste and upload paths into the app. Microsoft documents Endpoint DLP detection of pasting content to restricted service domains in supported browsers, and Purview DSPM for AI detects sensitive information shared with AI apps. Keep a dated record of the decision so an auditor can see when the feature was reviewed. Sources: [Endpoint data loss prevention, Microsoft Purview](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about); [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai). ## How do you keep the review current? Subscribe an owner to each large vendor's release notes and admin-center message feed, and review again when a vendor announces an AI feature. Add AI settings to the quarterly access review so a feature enabled by default is caught within a cycle. Nudge Security publishes a dated changelog of its own AI agent discovery coverage, which listed 23 agentic platforms on 2 November 2026; use dated vendor change logs like that to see which platforms a tool can read before relying on it. Sources: [Nudge Security changelog](https://www.nudgesecurity.com/changelog). ## Sources - [Nudge Security pricing](https://www.nudgesecurity.com/pricing) (nudgesecurity.com), read 2026-11-16 - [Grip Security: reduce shadow AI risks](https://www.grip.security/use-case-library/reduce-shadow-ai-risks) (grip.security), read 2026-11-16 - [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai) (help.zscaler.com), read 2026-11-16 - [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center) (docs.netskope.com), read 2026-11-16 - [Harmonic Security Explore](https://www.harmonic.security/products/explore) (harmonic.security), read 2026-11-16 - [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery) (reco.ai), read 2026-11-16 - [Grip Security: discover and control shadow AI](https://www.grip.security/use-case-library/discover-and-control-shadow-ai-across-your-enterprise) (grip.security), read 2026-11-16 - [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance) (learn.microsoft.com), read 2026-11-16 - [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps) (docs.paloaltonetworks.com), read 2026-11-16 - [Nudge Security: AI discovery solution](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution) (nudgesecurity.com), read 2026-11-16 - [Endpoint data loss prevention, Microsoft Purview](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about) (learn.microsoft.com), read 2026-11-16 - [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai) (learn.microsoft.com), read 2026-11-16 - [Nudge Security changelog](https://www.nudgesecurity.com/changelog) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Shadow AI Discovery From Identity Provider Logs](/guides/shadow-ai-discovery-from-identity-provider-logs/) - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/) - [embedded AI feature (glossary)](/glossary/embedded-ai-feature/) - [All guides](/guides/) - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # How to run a shadow AI audit URL: https://shadowaidetectiontools.com/guides/how-to-run-a-shadow-ai-audit/ Updated: 2026-11-16 A shadow AI audit lists every AI app, agent, extension, plugin and MCP server in use without approval, then rates each by data exposure and owner. Pull identity provider, proxy and endpoint data first, reconcile them into one inventory, and expect a first pass in days, not months. ## What does a shadow AI audit cover? A shadow AI audit covers seven kinds of unapproved AI: web apps, desktop apps, coding agents, browser extensions, IDE plugins, MCP servers and AI features switched on inside approved SaaS, plus apps granted OAuth access. Cover all of them, because each is visible from a different place and a single log source shows only some. Set the scope in writing before pulling data: which business units, which devices (managed and unmanaged), which identity tenants and which SaaS apps. Decide up front what counts as approved, since the audit output is the difference between what is in use and what is on the approved list. Sources: [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix); [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security). ## Which data sources do you pull first? Pull the identity provider logs first, then proxy or DNS logs, then SaaS admin consoles, then endpoint inventory. Identity records show which AI apps staff signed in to with a work account, and they exist before any new tool is bought. In Microsoft Entra ID the sign-in logs record every sign-in to the tenant, and the audit logs record application and consent changes. In Google Workspace, API controls list configured apps, accessed apps and apps pending review, with details appearing 24 to 48 hours after authorization. Okta exposes the System Log through an API. Then upload a firewall or proxy log: Microsoft documents a manual snapshot report for exactly this check before an automatic collector is set up. Sources: [Sign-in logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-sign-ins); [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs); [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227); [System Log API, Okta Developer](https://developer.okta.com/docs/reference/api/system-log/); [Create snapshot cloud discovery reports, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/create-snapshot-cloud-discovery-reports). ## How long does a first audit take? A first audit takes days when it relies on existing logs and about a week or more when new collection is needed. Published vendor setup times show the range: Grip Security says 10 minutes of setup and about 5 days to gather historic usage, Nudge Security says most organizations deploy in under an hour, and Harmonic says first insight in about 30 minutes. Reco publishes 24 to 48 hours to full deployment, and Microsoft documents that Purview DSPM for AI policies need at least 24 hours before results appear. All of these are vendor statements, so treat them as planning figures and measure your own. Sources: [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks); [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3); [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection); [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery); [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai); [Grip Security source (grip-security-14)](https://www.grip.security/ai-security); [Reco source (reco-08)](https://www.reco.ai/platform/application-discovery). ## What should the audit produce? The audit should produce one inventory with an owner for each entry, the data the tool touches, the sign-in method and a disposition: approve, restrict, replace or remove. A list of domains alone is not an inventory. Add the vantage point that found each item. An entry seen only in the proxy log may be a personal account, and an entry seen only in OAuth records may never have run on a managed device. The inventory guide on this site lists the fields. Sources: [View discovered apps on the Cloud discovery dashboard](https://learn.microsoft.com/en-us/defender-cloud-apps/discovered-apps); [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token). ## What should you do with the first findings? Triage by data exposure and by whether the account is personal or corporate, then pick one action per finding: warn the user, route to an approved tool, require approval, revoke the OAuth grant or block. Revoking an OAuth grant is a documented action in several products, and Microsoft documents reviewing and revoking permissions granted to enterprise applications in Entra ID. Schedule the next pass before closing the first. Cloud discovery policies can alert when a new app appears, which turns the audit into a standing control. Sources: [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions); [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies). ## Sources - [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix) (learn.microsoft.com), read 2026-11-16 - [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security) (harmonic.security), read 2026-11-16 - [Sign-in logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-sign-ins) (learn.microsoft.com), read 2026-11-16 - [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs) (learn.microsoft.com), read 2026-11-16 - [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227) (support.google.com), read 2026-11-16 - [System Log API, Okta Developer](https://developer.okta.com/docs/reference/api/system-log/) (developer.okta.com), read 2026-11-16 - [Create snapshot cloud discovery reports, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/create-snapshot-cloud-discovery-reports) (learn.microsoft.com), read 2026-11-16 - [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks) (grip.security), read 2026-11-16 - [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3) (nudgesecurity.com), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery) (reco.ai), read 2026-11-16 - [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai) (learn.microsoft.com), read 2026-11-16 - [View discovered apps on the Cloud discovery dashboard](https://learn.microsoft.com/en-us/defender-cloud-apps/discovered-apps) (learn.microsoft.com), read 2026-11-16 - [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token) (developers.google.com), read 2026-11-16 - [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions) (learn.microsoft.com), read 2026-11-16 - [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies) (learn.microsoft.com), read 2026-11-16 ## Related pages - [Triaging What a Shadow AI Scan Finds](/guides/triaging-what-a-shadow-ai-scan-finds/) - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/) - [AI asset register (glossary)](/glossary/ai-asset-register/) - [All guides](/guides/) - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Personal versus corporate AI accounts: telling them apart URL: https://shadowaidetectiontools.com/guides/personal-versus-corporate-ai-accounts/ Updated: 2026-11-16 A personal account is one signed in with a private email address; a corporate account is one in the company tenant. Browser extensions and identity records can tell them apart, tenant restrictions can enforce the split for some suites, and domain logs cannot. ## How do you tell a personal account from a work account? Read the account at sign-in, not the domain. A browser extension sees which account a person is logged in to, an identity provider logs which tenant issued the sign-in, and an API connection to an enterprise AI service lists corporate seats. A proxy log only shows that the host was contacted. Vendors document this signal. Akamai says its product can tell when a user reaches an AI tool with a personal rather than corporate account. Bay says its browser coverage detects corporate versus personal accounts, including personal Chrome profiles. Aurascape lists account type among its per-app attributes, and Harmonic says discovery covers personal accounts. Sources: [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control); [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Harmonic Security pricing](https://www.harmonic.security/pricing). ## What is tenant restriction? Tenant restriction is a control that limits sign-ins on your network to a list of permitted identity tenants and blocks all others. In Microsoft's implementation, users on the network can reach only the tenants an admin lists, and other tenants are blocked even where the user is a guest. Microsoft documents that a TLS-inspecting proxy inserts a header with the permitted tenant list into traffic to Entra ID, and that the feature covers apps that use Entra ID for single sign-on. It also supports blocking Microsoft consumer apps. It applies to Microsoft sign-ins, so it does not by itself separate personal and corporate accounts at other AI vendors. Sources: [Use tenant restrictions to manage access to SaaS apps](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/tenant-restrictions). ## Why does the difference matter for AI? Corporate AI accounts come with contract terms on data use and retention, admin controls and logs. Personal accounts usually carry consumer terms, and inputs may be retained or used for training depending on settings. The same prompt carries different risk in each. Per-tool terms vary, so record them. Nudge Security says it summarizes each AI vendor's training policy, opt-out options and retention, and Netskope says its risk scores include training on customer data. Sources: [Nudge Security: AI discovery solution](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution); [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center); [Netskope source (netskope-23)](https://www.netskope.com/products/securing-generative-ai). ## What are the ways to enforce the split? Four ways exist: tenant restriction for suites that support it, an enterprise license with SSO enforced, a browser control that warns or blocks personal logins, and a policy that routes users to an approved tool. Harmonic documents personal-account restriction, and Push Security documents warning with alternatives. Combine a control with a clear approved alternative. Blocking a personal account without a corporate option moves use to unmanaged devices. Sources: [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection); [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## What should the inventory record? Record the account type per user per tool, and report the share of use on personal accounts by tool. A falling share after an enterprise license rollout is a clean metric for the board. Keep the signal source with the number, since account detection from an extension and from an identity log can disagree. ## What do users need to hear? Users need a short rule and an easy alternative: use the company account for work, here is where to sign in, and here is whom to ask. A rule of this form can be delivered where the user is, at the moment a personal login is detected. Document the rule where the user will see it. Push Security describes warning users with alternatives, and Nudge Security describes delivering the AI acceptable use policy through its browser extension, which puts the text next to the action. Sources: [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [Nudge Security: browser extension](https://www.nudgesecurity.com/features/browser-extension). ## Sources - [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control) (akamai.com), read 2026-11-16 - [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Harmonic Security pricing](https://www.harmonic.security/pricing) (harmonic.security), read 2026-11-16 - [Use tenant restrictions to manage access to SaaS apps](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/tenant-restrictions) (learn.microsoft.com), read 2026-11-16 - [Nudge Security: AI discovery solution](https://www.nudgesecurity.com/try-nudge/ai-discovery-solution) (nudgesecurity.com), read 2026-11-16 - [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center) (docs.netskope.com), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 - [Nudge Security: browser extension](https://www.nudgesecurity.com/features/browser-extension) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/) - [tenant restriction (glossary)](/glossary/tenant-restriction/) - [personal account usage (glossary)](/glossary/personal-account-usage/) - [All guides](/guides/) - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Reading proxy and DNS logs for AI apps URL: https://shadowaidetectiontools.com/guides/reading-proxy-and-dns-logs-for-ai-apps/ Updated: 2026-11-16 Proxy and DNS logs show which AI domains devices contacted and when, which is enough to find web AI apps by name. They do not show the account used, the prompt sent, local agents, or AI inside approved apps on shared domains, so read them as one source among several. ## Which domains show AI use? Domains of AI web apps and API endpoints show AI use: chat sites, model provider APIs, image and transcription services, and the update or telemetry hosts of desktop assistants and coding agents. Build the watch list from a maintained catalog, not from memory, because new services appear weekly. Catalogs are the scalable route. Microsoft's cloud discovery compares traffic logs with a catalog of over 31,000 apps scored on more than 90 risk factors, Palo Alto Networks documents a catalog of GenAI apps, WitnessAI cites more than 4,000 AI apps, and Aurascape cites more than 20,000. Catalog sizes differ, so test with an AI tool you know is in use. Sources: [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery); [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps); [WitnessAI Observe](https://witness.ai/observe/); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Aurascape source (aurascape-07)](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/). ## What do proxy logs miss? Proxy logs miss the account, the content and anything that does not cross the proxy. Without TLS inspection a log records the destination host, not the path or the prompt, and a personal laptop off the VPN never appears. They also miss AI on shared domains. Microsoft explains in its tenant restrictions documentation that SaaS apps run on shared domain names such as outlook.office.com and login.microsoftonline.com, so blocking or watching the domain cannot separate one tenant or feature from another. The same limit applies to an AI assistant embedded in an approved suite. Sources: [Use tenant restrictions to manage access to SaaS apps](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/tenant-restrictions). ## How do you turn a raw log into an inventory? Normalize hostnames to registrable domains, join them to a catalog, count unique users and total bytes per app, and then rank by users and by upload volume. A single user with large uploads to an unknown AI host matters more than a hundred users on a known approved assistant. Microsoft documents a manual snapshot report that accepts firewall or proxy logs, parses them and shows app usage, which is a quick way to validate the log format before automating collection. A discovery policy can then alert when a new app appears. Sources: [Create snapshot cloud discovery reports, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/create-snapshot-cloud-discovery-reports); [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies). ## Where does DNS fit? DNS logs are broader than proxy logs because every device lookup passes the resolver, including unmanaged devices on corporate Wi-Fi and background lookups by agents. They show the name queried, the client and the time. They do not show URL paths, accounts or whether a query led to a connection. Use DNS to find devices and hosts you did not know about, then confirm the finding with a source that sees accounts or processes, such as an identity log or endpoint agent. Sources: [Zscaler: public AI inventory, endpoint AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-endpoint-ai); [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security). ## When should you add an inline or agent-based source? Add an inline source when you need action, not just a record: warn, redact or block at the moment of use. Inline products such as Aurascape, Netskope, Zscaler and WitnessAI document those actions, and WitnessAI states it works without an endpoint client or browser extension. Add an endpoint or browser source when the finding has to reach the account or the local tool. Cyberhaven, Harmonic and Bay document endpoint agents or browser extensions for that reason. Sources: [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center); [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai); [WitnessAI Observe](https://witness.ai/observe/); [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Aurascape source (aurascape-21)](https://aurascape.ai/frictionless-ai-security/); [Netskope source (netskope-18)](https://www.netskope.com/products/ai-products); [Zscaler source (zscaler-05)](https://www.zscaler.com/products-and-solutions/ai-security); [Harmonic Security source (harmonic-security-21)](https://www.harmonic.security/products/command); [Bay source (bay-14)](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Bay source (bay-12)](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide). ## How often should the domain list be refreshed? Refresh a hand-built list weekly and prefer a catalog the product updates for you. New AI services appear faster than a spreadsheet can track, and an out-of-date list produces a clean report that is wrong. Check how a vendor adds new apps. Aurascape lists a 48-hour service level for connectors to new AI apps, and Microsoft tells customers to check its roadmap or create a custom app for anything not in the catalog. Sources: [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery); [Aurascape: AI application discovery](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/). ## Sources - [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery) (learn.microsoft.com), read 2026-11-16 - [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps) (docs.paloaltonetworks.com), read 2026-11-16 - [WitnessAI Observe](https://witness.ai/observe/) (witness.ai), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Use tenant restrictions to manage access to SaaS apps](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/tenant-restrictions) (learn.microsoft.com), read 2026-11-16 - [Create snapshot cloud discovery reports, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/create-snapshot-cloud-discovery-reports) (learn.microsoft.com), read 2026-11-16 - [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies) (learn.microsoft.com), read 2026-11-16 - [Zscaler: public AI inventory, endpoint AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-endpoint-ai) (help.zscaler.com), read 2026-11-16 - [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security) (harmonic.security), read 2026-11-16 - [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center) (docs.netskope.com), read 2026-11-16 - [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai) (help.zscaler.com), read 2026-11-16 - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 ## Related pages - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [DNS-based detection (glossary)](/glossary/dns-based-detection/) - [proxy log analysis (glossary)](/glossary/proxy-log-analysis/) - [All guides](/guides/) - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI detection evidence for auditors URL: https://shadowaidetectiontools.com/guides/shadow-ai-detection-evidence-for-auditors/ Updated: 2026-11-16 Auditors ask for a dated inventory of AI tools, the sources it came from, who approved each tool, and records showing the control ran. Keep detection logs for the period your audit scope requires, and note that platform defaults differ: Microsoft Audit (Standard) keeps records 180 days. ## What evidence do auditors ask for? Auditors ask for five things: the inventory with dates, the discovery method and sources, the approval record per tool, proof the control operated over the period, and exceptions with sign-off. A screenshot of a dashboard shows one day; a dated export shows the period. Pick tools that export. Harmonic's documentation describes CSV export of the per-app risk profile, Zscaler documents an Excel export for its MCP inventory, and for several products an inventory export was not found on 16 Nov 2026 in the vendor files. Sources: [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection); [Zscaler: AI assets, MCP servers](https://help.zscaler.com/ai-asset-mgmt/about-ai-assets-mcp-servers); [Harmonic Security source (harmonic-security-36)](https://docs.harmonicsecurity.app/llms.txt). ## How long should detection logs be kept? Keep them for the period set by your audit scope or regulation, and confirm what the platform keeps by default. Microsoft Purview Audit (Standard) retains audit records for 180 days, and the page notes that longer retention requires other options, such as exporting through the Office 365 Management Activity API to a SIEM. Cloud discovery and endpoint tools have their own retention, so ask each vendor. The answer belongs in the evidence file next to the export schedule. Sources: [Auditing solutions in Microsoft Purview](https://learn.microsoft.com/en-us/purview/audit-solutions-overview). ## How do you show the control actually ran? Show scheduled runs, alerts raised, and dispositions closed. Cloud discovery policies that alert on new apps leave a record of alerts, and identity logs show consent grants and revocations by date. Entra ID audit logs capture changes to applications, groups, users and licenses. Pair each alert with the ticket that closed it. Several products route to a ticket: Push Security documents Jira tickets through a webhook, and Cyberhaven documents ServiceNow and SOAR integrations. Sources: [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies); [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs); [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Push Security help: sending Push data to a SIEM or ticketing system](https://pushsecurity.com/help/can-i-send-push-data-to-other-systems-like-a-siem-or-ticketing-system); [Cyberhaven integrations](https://www.cyberhaven.com/product/integrations). ## What about risk assessments of each tool? Keep the assessment on file per approved tool: data processed, training and retention terms, hosting, certifications and the decision date. Vendors provide starting data. Microsoft's catalog scores more than 31,000 apps on 90-plus risk factors, Nudge Security offers over 200,000 SaaS vendor security profiles, and Zscaler documents a risk index with hosting, posture and retention. A vendor score is input, not the assessment. The reviewer's decision and date are the evidence. Sources: [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery); [Nudge Security](https://www.nudgesecurity.com/); [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai). ## What does a minimal evidence pack look like? A minimal pack holds the policy, the scope statement, the dated inventory exports, the source list, the approval records, alert and ticket samples, and the exception register. Store it where the auditor can read it without asking the tool owner. Assemble it quarterly so no audit starts from zero. ## What gaps do auditors find most often? The common gaps are an inventory with no dates, no stated scope, no owner on entries, and exceptions without sign-off. A second gap is relying on one source: an auditor can ask which devices and tenants were covered, and the answer must be a number. Fix them by adding the coverage line to the inventory header, storing the policy version with the export, and keeping exceptions in a register with an expiry date. None of these needs a new tool. If a tool is added mid-period, record the date it went live. An auditor reading a count that rises sharply will look for the reason, and a dated source list explains it before the question is asked. Sources: [Auditing solutions in Microsoft Purview](https://learn.microsoft.com/en-us/purview/audit-solutions-overview). ## Sources - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Zscaler: AI assets, MCP servers](https://help.zscaler.com/ai-asset-mgmt/about-ai-assets-mcp-servers) (help.zscaler.com), read 2026-11-16 - [Auditing solutions in Microsoft Purview](https://learn.microsoft.com/en-us/purview/audit-solutions-overview) (learn.microsoft.com), read 2026-11-16 - [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies) (learn.microsoft.com), read 2026-11-16 - [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs) (learn.microsoft.com), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery) (learn.microsoft.com), read 2026-11-16 - [Nudge Security](https://www.nudgesecurity.com/) (nudgesecurity.com), read 2026-11-16 - [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai) (help.zscaler.com), read 2026-11-16 ## Related pages - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/) - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/) - [AI asset register (glossary)](/glossary/ai-asset-register/) - [All guides](/guides/) - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [Shadow AI Detection in Microsoft 365 and Google Workspace](/guides/shadow-ai-detection-in-microsoft-365-and-google-workspace/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI detection in Microsoft 365 and Google Workspace URL: https://shadowaidetectiontools.com/guides/shadow-ai-detection-in-microsoft-365-and-google-workspace/ Updated: 2026-11-16 Microsoft Defender for Cloud Apps, Defender for Endpoint and Purview cover discovery, local agents and paste control in Microsoft 365, with gaps for uncatalogued apps. Google Workspace admin logs show OAuth app access and token events. Neither shows personal-account use on unmanaged devices. ## What can Microsoft Purview and Defender for Cloud Apps see? Defender for Cloud Apps discovers cloud apps from three feeds: the Defender for Endpoint integration on Windows 10 and 11, a log collector on firewalls and proxies, and proxy integrations. It compares traffic to a catalog of over 31,000 apps and lets admins unsanction an app. Purview adds AI sites seen through browser activity, classified as Generative AI in the catalog. Third-party AI site insights in DSPM for AI need the Purview browser extension and onboarded devices, and policies need at least 24 hours to collect data. Endpoint DLP detects pasting content to restricted service domains in supported browsers. Microsoft's own tutorial says apps outside the catalog cannot be discovered by default. Sources: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Microsoft Purview and AI](https://learn.microsoft.com/en-us/purview/ai-microsoft-purview); [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai); [Endpoint data loss prevention, Microsoft Purview](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about). ## What does Defender for Endpoint add for local AI agents? Defender for Endpoint discovers supported local AI agents on Windows devices, including coding agents, IDE extensions, desktop assistants and agent platforms, with no extra deployment beyond onboarding. Microsoft's support matrix names Claude Code, Claude Desktop, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf and Kiro. The release notes of October 2026 say discovery is generally available on Windows and macOS and reports each agent's configured MCP servers. Agent risk levels need Microsoft 365 E7 or Agent 365 with Defender for Endpoint Plan 2. Sources: [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents); [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix); [What's new in Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint). ## What can Google Workspace admin logs see? Google Workspace admin logs see OAuth app access, not browser traffic. API controls list configured apps, accessed apps and apps pending review, with access levels of Trusted, Limited, Specific Google data or Blocked, and details appear 24 to 48 hours after authorization. The Reports API records OAuth token audit events, and the OAuth log events page in Admin Help describes them. Use these to find AI apps that were granted access to Gmail, Drive or Calendar, and to block or limit them. Sources: [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227); [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token); [OAuth log events, Google Workspace Admin Help](https://support.google.com/a/answer/6124308). ## What do both suites miss? Both miss personal accounts on unmanaged devices and AI tools that never authenticate through the suite. Microsoft's catalog gap applies to new AI tools, and neither suite's admin logs show the content sent to a third-party AI app. Add a browser or endpoint source and an identity or SaaS API source from a dedicated product. The Vantage Point Matrix on this site compares them. Sources: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it). ## What should a Microsoft 365 shop do first? Start with what is already licensed: confirm Defender for Endpoint onboarding for local agent discovery, set up cloud discovery from the firewall or proxy, review app governance for OAuth apps, and turn on user consent restrictions. For a Google Workspace shop, start with API controls: review accessed apps, set the unknown ones to Limited or Blocked, and watch the OAuth token events. Sources: [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance); [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent); [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227). ## How do you prove coverage in either suite? Count the devices onboarded to Defender for Endpoint against the device list, and the apps with a recorded access level in Google Workspace API controls against the apps that have accessed data. Those two ratios show how much of the estate the suite can see. Microsoft lists Windows, macOS, Linux, Android and iOS as Defender for Endpoint operating systems, while local agent discovery is documented for Windows and macOS, so state which devices the number covers. Sources: [Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/microsoft-defender-endpoint); [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227). ## Sources - [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it) (learn.microsoft.com), read 2026-11-16 - [Microsoft Purview and AI](https://learn.microsoft.com/en-us/purview/ai-microsoft-purview) (learn.microsoft.com), read 2026-11-16 - [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai) (learn.microsoft.com), read 2026-11-16 - [Endpoint data loss prevention, Microsoft Purview](https://learn.microsoft.com/en-us/purview/endpoint-dlp-learn-about) (learn.microsoft.com), read 2026-11-16 - [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents) (learn.microsoft.com), read 2026-11-16 - [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix) (learn.microsoft.com), read 2026-11-16 - [What's new in Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/whats-new-in-microsoft-defender-endpoint) (learn.microsoft.com), read 2026-11-16 - [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227) (support.google.com), read 2026-11-16 - [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token) (developers.google.com), read 2026-11-16 - [OAuth log events, Google Workspace Admin Help](https://support.google.com/a/answer/6124308) (support.google.com), read 2026-11-16 - [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance) (learn.microsoft.com), read 2026-11-16 - [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent) (learn.microsoft.com), read 2026-11-16 - [Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/microsoft-defender-endpoint) (learn.microsoft.com), read 2026-11-16 ## Related pages - [Shadow AI Discovery From Identity Provider Logs](/guides/shadow-ai-discovery-from-identity-provider-logs/) - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/) - [CASB (glossary)](/glossary/casb/) - [All guides](/guides/) - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI discovery from identity provider logs URL: https://shadowaidetectiontools.com/guides/shadow-ai-discovery-from-identity-provider-logs/ Updated: 2026-11-16 Identity provider logs show which AI apps staff signed in to with a work account and which AI apps were granted OAuth access. Read sign-in, consent and app-grant events in Entra ID, Google Workspace or Okta; they miss tools used with personal accounts and anything that never calls the identity provider. ## What does single sign-on reveal about AI use? Single sign-on reveals every AI app a work identity signed in to through the identity provider, with the user, the time, the app and the result. Microsoft says Entra ID logs all sign-ins into a tenant, so an AI app that uses "Sign in with Microsoft" appears there even if no one registered it. It reveals less about AI used with a personal email address, a password typed into the app, or a tool that never authenticates. Grip Security, Reco and Nudge Security build discovery on identity and email sources. Grip Security and Nudge Security document named identity provider integrations such as Okta, Entra ID and Google Workspace, and Reco says its discovery draws on SSO logs. Sources: [Sign-in logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-sign-ins); [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks); [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3); [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery); [Grip Security source (grip-security-11)](https://www.grip.security/integrations); [Reco source (reco-06)](https://www.reco.ai/platform/application-discovery). ## Which sign-in events matter? Three event types matter: first-time sign-ins to an application, consent grants to a new application, and permission changes on an existing grant. Entra ID separates interactive and non-interactive user sign-ins, service principal sign-ins and managed identity sign-ins, and its audit log records changes to applications and consent. In Google Workspace, the OAuth log events and the token audit activity events in the Reports API record authorization, revocation and API requests by app. In Okta, the System Log API gives near real-time, read-only access to events as an audit trail. Filter each for application names and client IDs you do not recognize, and group by user count. Sources: [Sign-in logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-sign-ins); [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs); [OAuth log events, Google Workspace Admin Help](https://support.google.com/a/answer/6124308); [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token); [System Log API, Okta Developer](https://developer.okta.com/docs/reference/api/system-log/). ## How do OAuth grants show AI apps with access to data? An OAuth grant lists the app, the user or admin who consented and the permissions given. AI note takers, writing assistants and chat tools often request mailbox, calendar or file permissions, so the grant list is the quickest way to see which of them can read company data. Microsoft's app governance covers OAuth-enabled apps registered in Entra ID, Google and Salesforce, showing the permissions they hold and who granted access. Google Workspace API controls list configured apps, accessed apps and apps pending review, with a Trusted, Limited, Specific Google data or Blocked setting. Sources: [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance); [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227). ## What controls follow from what you find? Restrict who may consent, require an approval path, and revoke grants that are risky. Entra ID lets admins limit user consent and turn on an admin consent workflow, so a user request goes to a reviewer. Microsoft also documents reviewing and revoking permissions granted to enterprise applications. Several detection products automate part of this: Nudge Security lists auto-revoking overly permissive OAuth grants (a paid-tier item), and Grip says it categorizes OAuth scopes by risk and can revoke access. Sources: [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent); [Overview of admin consent workflow](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/admin-consent-workflow-overview); [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions); [Nudge Security pricing](https://www.nudgesecurity.com/pricing); [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks). ## What do identity logs not show? Identity logs do not show local tools, so a coding agent, a local model or an MCP server on a laptop never appears. They do not show the content sent to an AI app, and they miss AI features switched on inside an app that was already approved, because no new sign-in occurs. Pair identity logs with an endpoint or browser source for those cases. Harmonic documents an endpoint agent for local apps, coding assistants and MCP servers, and Push Security documents a browser extension that sees logins to apps, including AI apps. Sources: [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security); [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## Sources - [Sign-in logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-sign-ins) (learn.microsoft.com), read 2026-11-16 - [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks) (grip.security), read 2026-11-16 - [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3) (nudgesecurity.com), read 2026-11-16 - [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery) (reco.ai), read 2026-11-16 - [Audit logs in Microsoft Entra ID](https://learn.microsoft.com/en-us/entra/identity/monitoring-health/concept-audit-logs) (learn.microsoft.com), read 2026-11-16 - [OAuth log events, Google Workspace Admin Help](https://support.google.com/a/answer/6124308) (support.google.com), read 2026-11-16 - [OAuth Token Audit Activity Events, Google for Developers](https://developers.google.com/workspace/admin/reports/v1/appendix/activity/token) (developers.google.com), read 2026-11-16 - [System Log API, Okta Developer](https://developer.okta.com/docs/reference/api/system-log/) (developer.okta.com), read 2026-11-16 - [Manage app governance, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/app-governance-manage-app-governance) (learn.microsoft.com), read 2026-11-16 - [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227) (support.google.com), read 2026-11-16 - [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent) (learn.microsoft.com), read 2026-11-16 - [Overview of admin consent workflow](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/admin-consent-workflow-overview) (learn.microsoft.com), read 2026-11-16 - [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions) (learn.microsoft.com), read 2026-11-16 - [Nudge Security pricing](https://www.nudgesecurity.com/pricing) (nudgesecurity.com), read 2026-11-16 - [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security) (harmonic.security), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 ## Related pages - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [OAuth grant (glossary)](/glossary/oauth-grant/) - [All guides](/guides/) - [Shadow AI Detection in Microsoft 365 and Google Workspace](/guides/shadow-ai-detection-in-microsoft-365-and-google-workspace/) - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI false positives and blind spots URL: https://shadowaidetectiontools.com/guides/shadow-ai-false-positives-and-blind-spots/ Updated: 2026-11-16 Detection tools get three things wrong: they flag approved tools under new domains, they miss apps outside their catalog, and they cannot see use on unmanaged devices or personal accounts. The vantage point decides what stays invisible: network tools miss local agents, endpoint tools miss SaaS grants. ## What do detection tools get wrong? They get catalog matches wrong in both directions. A new domain for an approved tool is flagged as new, and a new AI tool outside the catalog is missed. Microsoft states that apps outside the Defender for Cloud Apps catalog cannot be discovered by default. They also mislabel the account. A domain log cannot tell a personal from a corporate login, so a tool that counts users by domain can overstate corporate use. Products that read the browser or identity record can separate them, and Harmonic and Bay both describe personal-account detection. Sources: [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery); [Harmonic Security pricing](https://www.harmonic.security/pricing); [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Harmonic Security source (harmonic-security-29)](https://www.harmonic.security/solutions/shadow-ai-detection). ## Which kinds of AI use stay invisible? Five stay invisible to some tools. Use on an unmanaged phone or home laptop reaches no corporate sensor. Use inside an approved app's new AI feature leaves no new domain. Local models and CLI agents leave no network trace. Pasted text inside a permitted AI app is visible only to tools that inspect content. Offline use leaves nothing. Each of those has a matching vantage point: a browser extension for personal-account web use, SaaS APIs for embedded features, an endpoint agent for local models and CLI agents, and content inspection for pasted text. Sources: [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security); [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai); [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## Where does each vantage point go blind? Network and proxy tools are blind to local software and to devices off the network. Browser-only tools are blind to desktop apps and CLI agents. Identity and email tools are blind to anything that never authenticates through the provider. Endpoint agents are blind to unmanaged devices. Product files bear this out: Nudge Security says it does not touch the corporate network or endpoints, WitnessAI is network-only, and Push Security's vantage is the browser. Choose a second source that covers the first one's blind spot. Sources: [Nudge Security product](https://www.nudgesecurity.com/product); [WitnessAI Observe](https://witness.ai/observe/); [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## How do you test a tool's coverage? Plant known cases and see which are found: a test AI web app on a managed laptop, a personal-account login in a browser, an extension, a coding agent with an MCP server, and an OAuth grant on a test user. Record which tool found which and how long it took. Run the test in a proof of value, with the same cases for every candidate. Vendors state times from minutes to 48 hours; your cases show the real figure. Sources: [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery); [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection). ## How should the limits be reported? Report limits next to the numbers: the sources used, the devices covered and the kinds not covered. A board count of "37 AI tools" without a coverage line suggests completeness the data cannot support. State the share of managed devices with an agent, and the identity tenants connected. ## Which fixes reduce false alarms most? Three fixes cut false alarms the most: maintain an approved list so known tools are matched, tag corporate IP ranges and test accounts, and connect an identity source so sign-ins resolve to people and tenants. The first two are policy work; the third needs an integration. Check that the product separates the approved version of a tool from the consumer version. The same brand can appear as a licensed enterprise tenant and a personal login, and only a source that reads the account tells them apart. Sources: [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control). ## Sources - [Cloud app discovery overview, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/set-up-cloud-discovery) (learn.microsoft.com), read 2026-11-16 - [Harmonic Security pricing](https://www.harmonic.security/pricing) (harmonic.security), read 2026-11-16 - [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security) (harmonic.security), read 2026-11-16 - [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai) (help.zscaler.com), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 - [Nudge Security product](https://www.nudgesecurity.com/product) (nudgesecurity.com), read 2026-11-16 - [WitnessAI Observe](https://witness.ai/observe/) (witness.ai), read 2026-11-16 - [Reco: shadow AI discovery](https://www.reco.ai/use-cases/shadow-ai-discovery) (reco.ai), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control) (akamai.com), read 2026-11-16 ## Related pages - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [traffic fingerprinting (glossary)](/glossary/traffic-fingerprinting/) - [All guides](/guides/) - [Shadow AI Discovery From Identity Provider Logs](/guides/shadow-ai-discovery-from-identity-provider-logs/) - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI metrics for the board URL: https://shadowaidetectiontools.com/guides/shadow-ai-metrics-for-the-board/ Updated: 2026-11-16 A board report on shadow AI needs six numbers: tools found, active users, share of use on personal accounts, high-risk findings open, median time to resolve, and coverage. Report monthly during a rollout and quarterly after, and state which sources and devices the numbers cover. ## Which numbers belong in a board report? Six belong: AI tools in use found, active users per tool, share of use on personal accounts, high-risk findings open, median days to resolve a finding, and coverage (devices and identity tenants with a sensor). Each answers a board question: how much, who, how exposed, how fast, and how sure. The data exists in the products. Microsoft's cloud discovery dashboard shows the types of apps used, open alerts, risk levels and top users. Zscaler documents usage by user and department, and Cyberhaven records who uses each tool and with what data. Sources: [View discovered apps on the Cloud discovery dashboard](https://learn.microsoft.com/en-us/defender-cloud-apps/discovered-apps); [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai); [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Zscaler source (zscaler-12)](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-overview); [Cyberhaven source (cyberhaven-23)](https://www.cyberhaven.com/technologies/gen-ai). ## How often should they be reported? Report monthly while the program is being rolled out and quarterly once it is steady, with an immediate note for a serious event. The monthly view shows whether discovery coverage is rising, and the quarterly view shows trend. Align with the audit cycle so the board pack is the same evidence the auditor reads. ## How do you present a first inventory fast? For a first number inside 30 days, use whichever source can be live soonest and say what it covers. Vendors publish setup times such as Grip Security's 10 minutes with about 5 days for historic usage, Harmonic's first insight in about 30 minutes, and Bay's claim of under 5 minutes via one MDM or EDR script. Present the count together with the coverage line, and promise a second source by a named date. Sources: [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks); [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection); [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Grip Security source (grip-security-14)](https://www.grip.security/ai-security); [Bay source (bay-18)](https://www.bay.io/platform/ai-endpoint-security?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide). ## What should be left out? Leave out raw domain counts, vendor risk scores shown without context, and catalog size. A tool catalog of 4,000 or 20,000 apps says how much a product recognizes, not how much your staff use. Leave out any figure you cannot reproduce from a source. Do not present a vendor's own benchmark of AI use as your own organization's data. Sources: [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/). ## How do you show progress, not just exposure? Show before and after for each action: personal-account share, tools moved to approved alternatives, grants revoked and blocks enforced. Progress numbers keep the board from reading a rising tool count as a failing program, since better discovery finds more. State that explicitly: a higher count after a new source is added reflects coverage. Sources: [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps). ## What should the chart show? Use one trend line for tools found and one for coverage, plotted on the same dates. When discovery improves, the count of tools rises first and then flattens as the inventory catches up; the coverage line explains the step. Add a small table of the five highest-risk findings with owner, action and date. The board reads names and dates faster than totals, and it shows who is accountable for each. Label every number with its source and date. A vendor dashboard figure is a vendor-computed value, and the report should say which product produced it. Microsoft's cloud discovery dashboard, for example, shows app types, open alerts, risk levels and top users, and a board slide quoting it should say so. Close with the decisions you need from the board, such as approving an enterprise license for a tool staff already use. Keep that to a line or two. Sources: [View discovered apps on the Cloud discovery dashboard](https://learn.microsoft.com/en-us/defender-cloud-apps/discovered-apps). ## Sources - [View discovered apps on the Cloud discovery dashboard](https://learn.microsoft.com/en-us/defender-cloud-apps/discovered-apps) (learn.microsoft.com), read 2026-11-16 - [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai) (help.zscaler.com), read 2026-11-16 - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks) (grip.security), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps) (docs.paloaltonetworks.com), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 ## Related pages - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [usage heatmap (glossary)](/glossary/usage-heatmap/) - [All guides](/guides/) - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/) - [Shadow AI vs Shadow IT: What Changes for Detection](/guides/shadow-ai-vs-shadow-it-what-changes-for-detection/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI vs shadow IT: what changes for detection URL: https://shadowaidetectiontools.com/guides/shadow-ai-vs-shadow-it-what-changes-for-detection/ Updated: 2026-11-16 Shadow IT detection looks for new apps in network traffic against a catalog. Shadow AI adds local agents, extensions, MCP servers and AI features that appear inside approved apps, so catalog-and-traffic discovery alone misses part of it. Detection needs endpoint, identity and SaaS sources as well. ## How is shadow AI different from shadow IT? Shadow IT is an unapproved app or service, usually found as a new domain in traffic. Shadow AI includes that, and also covers things with no new domain: a coding agent installed on a laptop, a browser extension, an MCP server in a config file, or an AI feature a vendor switches on inside an approved SaaS app. Detection therefore moves from one vantage point to several. Microsoft's own shadow IT tutorial describes discovery from network traffic compared with an app catalog, and Harmonic's endpoint page lists desktop apps, coding assistants, command-line tools, local models and MCP servers as separate finds. Sources: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security). ## Why do old SaaS discovery methods miss AI? Catalog-based methods miss AI that is new, local or embedded. Microsoft states in its tutorial that apps outside the catalog cannot be discovered by default, which matters because new AI tools appear faster than any catalog can add them. Traffic-only methods also miss anything that runs on the device. Zscaler documents three separate inventories (GenAI web apps, embedded AI and endpoint AI) because one traffic view does not cover all three, and Aurascape describes a long tail of AI apps beyond a standard catalog. Sources: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai); [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai); [Zscaler: public AI inventory, endpoint AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-endpoint-ai); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/). ## What does shadow AI add to the risk? Shadow AI adds data exposure at the prompt. A shadow IT app stores files; an AI app receives text, code and records typed or pasted into it, and some use inputs for training. Several products therefore score training and retention per app: Aurascape lists training and retention among 25-plus attributes, and Netskope includes training on customer data in its risk scores. It also adds autonomy. A coding agent or MCP-connected assistant can act on files and services, which is why Cyberhaven and Bay inventory agent frameworks and MCP servers, not only apps. Sources: [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center); [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Netskope source (netskope-23)](https://www.netskope.com/products/securing-generative-ai); [Aurascape source (aurascape-28)](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/). ## Which parts of the shadow IT toolkit still apply? Identity and consent controls still apply. Entra ID lets admins restrict user consent and enable an admin consent workflow, and Google Workspace API controls let admins mark apps trusted, limited or blocked. Both limit AI apps that ask for OAuth access just as they limit any other app. Cloud discovery policies, firewall log analysis and tenant restrictions also carry over. They remain the cheapest first layer, as long as the audit adds an endpoint or browser source for what they cannot see. Sources: [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent); [Overview of admin consent workflow](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/admin-consent-workflow-overview); [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227); [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies). ## How should a program combine the methods? Run at least one identity or SaaS API source, one traffic source and one endpoint or browser source. The Vantage Point Matrix on this site shows which products read which sources, and vendor profiles name the sources behind each cell. Reconcile the three into a single inventory and record which source saw each item. Disagreement between sources is itself a finding, for instance a tool seen in identity records but never on a managed device. Sources: [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3); [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps); [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection). ## What should a security lead ask a vendor? Ask which kinds of AI the product finds and from which source, and ask to see the evidence for each kind. A product that documents web apps from traffic may say nothing about coding agents or MCP servers, and the answer should name the vantage point, such as endpoint agent, browser extension, proxy, identity provider or SaaS API. Ask also how the product handles an app outside its catalog, how quickly a new AI service is added (Aurascape lists a 48-hour service level for connectors to new AI apps), and what it does when it finds one. The ranking on this site scores each of these questions from published evidence, and each vendor profile lists the facts. Sources: [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it); [Aurascape: AI application discovery](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/). ## Sources - [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it) (learn.microsoft.com), read 2026-11-16 - [Harmonic Security: endpoint AI security](https://www.harmonic.security/solutions/endpoint-ai-security) (harmonic.security), read 2026-11-16 - [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai) (help.zscaler.com), read 2026-11-16 - [Zscaler: public AI inventory, embedded AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-embedded-ai) (help.zscaler.com), read 2026-11-16 - [Zscaler: public AI inventory, endpoint AI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-endpoint-ai) (help.zscaler.com), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center) (docs.netskope.com), read 2026-11-16 - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Configure how users consent to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent) (learn.microsoft.com), read 2026-11-16 - [Overview of admin consent workflow](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/admin-consent-workflow-overview) (learn.microsoft.com), read 2026-11-16 - [Control which apps access Google Workspace data](https://support.google.com/a/answer/7281227) (support.google.com), read 2026-11-16 - [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies) (learn.microsoft.com), read 2026-11-16 - [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3) (nudgesecurity.com), read 2026-11-16 - [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps) (docs.paloaltonetworks.com), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 ## Related pages - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [shadow IT (glossary)](/glossary/shadow-it/) - [All guides](/guides/) - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/) - [Triaging What a Shadow AI Scan Finds](/guides/triaging-what-a-shadow-ai-scan-finds/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Triaging what a shadow AI scan finds URL: https://shadowaidetectiontools.com/guides/triaging-what-a-shadow-ai-scan-finds/ Updated: 2026-11-16 Triage a shadow AI scan by what the tool can reach, whose account it uses and whether it can act on its own. Urgent findings are personal accounts holding sensitive data, OAuth grants with broad mail or file scopes, and unvetted MCP servers on developer machines. ## How do you rank findings by risk? Rank by four factors: data reached, account type, access scope and autonomy. Data is what staff paste or connect. Account type separates corporate tenants with contract terms from personal accounts with none. Scope is the permission an OAuth grant holds. Autonomy is whether the tool acts on its own. Products score some of these for you. Bay says it gives each tool a 1-to-5 autonomy score and scans for malicious indicators, Netskope's risk scores include training on customer data, and Cyberhaven scores each app on five dimensions. Treat the vendor score as a starting point and check it against your data classes. Sources: [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center); [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Netskope source (netskope-23)](https://www.netskope.com/products/securing-generative-ai). ## Which findings are urgent? Three categories are urgent: sensitive data going to a personal-account AI app, OAuth grants with broad mailbox or file permissions held by an app nobody reviewed, and MCP servers or coding agents on developer machines that no one vetted. For the OAuth case, Entra ID's permission review lists what an app holds and lets an admin revoke it. For personal accounts, Akamai says its product can tell when a user reaches an AI tool with a personal rather than corporate account, and Bay says its browser coverage identifies accounts. For MCP servers, Microsoft reports each discovered agent's configured MCP servers. Sources: [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions); [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control); [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide); [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents). ## What action goes with each finding? Match one action to each finding: warn or coach when the tool is low risk and an approved alternative exists, require approval for a new but plausible tool, redact or stop sensitive data, revoke an OAuth grant, or block when the tool is malicious or the data is regulated. Documented examples: Push Security can warn with alternatives and redact on paste, Aurascape documents time-limited exception requests, and Palo Alto Networks documents revoking access by privilege scope. The Finding-to-Action Ladder on this site shows which products document each action. Sources: [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps); [Aurascape source (aurascape-22)](https://aurascape.ai/frictionless-ai-security/); [Palo Alto Networks source (palo-alto-networks-18)](https://www.paloaltonetworks.com/sase/ai-access-security). ## How should false alarms be handled? Mark confirmed false positives with a reason and a date, so the same finding is not reopened. A false alarm is often an approved tool under a new domain or a developer's test install. Keep the entry in the inventory with a disposition of none, not a deletion. If a tool reports many false alarms for one source, check the catalog match before adjusting policy. ## When is a finding closed? A finding is closed when its disposition is recorded and the control is in place: the app is approved and owned, the user is moved to an approved tool, the grant is revoked, or the block is enforced. Re-scan after the action to confirm the item no longer appears. Report the counts by disposition to the owner each month. ## Who should be told, and when? Tell the tool's business owner within a working day for urgent findings, and batch the rest into a weekly list. The message names the tool, the risk, the approved alternative and the deadline, and avoids blame. People who feel punished stop using company devices for AI and move to personal ones, where detection ends. Where the product can speak to the user directly, use it. Harmonic documents inline warn and nudge messages, and Nudge Security documents nudging staff toward sanctioned tools and delivering the AI acceptable use policy through its browser extension. Sources: [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection); [Nudge Security: browser extension](https://www.nudgesecurity.com/features/browser-extension); [Harmonic Security source (harmonic-security-28)](https://www.harmonic.security/pricing). ## Sources - [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Netskope AI Command Center documentation](https://docs.netskope.com/en/ai-command-center) (docs.netskope.com), read 2026-11-16 - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions) (learn.microsoft.com), read 2026-11-16 - [Akamai: AI usage control](https://www.akamai.com/solutions/ai-usage-control) (akamai.com), read 2026-11-16 - [Bay: browser platform](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=guide) (bay.io), read 2026-11-16 - [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents) (learn.microsoft.com), read 2026-11-16 - [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps) (docs.paloaltonetworks.com), read 2026-11-16 - [Harmonic Security: shadow AI detection](https://www.harmonic.security/solutions/shadow-ai-detection) (harmonic.security), read 2026-11-16 - [Nudge Security: browser extension](https://www.nudgesecurity.com/features/browser-extension) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/) - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/) - [AI risk rating (glossary)](/glossary/ai-risk-rating/) - [All guides](/guides/) - [Shadow AI vs Shadow IT: What Changes for Detection](/guides/shadow-ai-vs-shadow-it-what-changes-for-detection/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # What a shadow AI inventory should contain URL: https://shadowaidetectiontools.com/guides/what-a-shadow-ai-inventory-should-contain/ Updated: 2026-11-16 A shadow AI inventory needs one row per tool with the kind, owner, users, data reached, account type, source that saw it, risk, decision and review date. An owner for every entry matters most, because an unowned row cannot be approved, restricted or removed. ## Which fields belong in the inventory? Nine fields do the work: tool name and vendor, kind (web app, desktop app, coding agent, extension, IDE plugin, MCP server, embedded feature, OAuth app), number and names of users, data reached, account type (corporate or personal), the source that saw it, risk, disposition and review date. Vendors publish similar sets. Cyberhaven says it scores each app on five risk dimensions and records who uses each tool and with what data. Zscaler describes a per-app risk index with hosting, posture and retention, and usage by user and department. Aurascape lists more than 25 risk attributes per app, including account type. Sources: [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security); [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai); [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/); [Aurascape source (aurascape-28)](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/). ## Who owns each entry? Name one accountable owner per entry: the business owner who wants the tool, not the security team that found it. Security owns the review process, the business owner owns the decision and the data. Where users are many and no owner exists, assign the department head with the most users. Nudge Security lets admins set an approval status of Approved, Allowed, In Review or Not Permitted for each discovered agent, a pattern worth copying in any spreadsheet. Sources: [Nudge Security: AI agent discovery](https://www.nudgesecurity.com/features/ai-agent-discovery). ## How do you record the evidence behind an entry? Record the source and the date each entry was first and last seen. An auditor will ask how you know, and the answer is a log, an API record or an endpoint scan with a date. Keep the query or export that produced it. Microsoft's Purview audit retention is 180 days for Audit (Standard), and records older than that are gone unless exported. Plan to export inventory snapshots on a schedule so history survives. Sources: [Auditing solutions in Microsoft Purview](https://learn.microsoft.com/en-us/purview/audit-solutions-overview). ## How should kinds of AI be tracked separately? Track kinds separately because the control differs. A web app is blocked or coached at the browser or proxy, a coding agent is governed on the device, an MCP server is approved by name and version, and an OAuth app is revoked in the identity provider. Microsoft's Defender for Endpoint support matrix lists local agents by name, including Claude Code, Cursor, Codex CLI, Gemini CLI and GitHub Copilot, and it reports each agent's configured MCP servers once discovery is available. Sources: [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix); [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents). ## How often should it be refreshed? Refresh continuously from automated sources and review the full inventory at least quarterly. New AI tools appear weekly, and a quarterly cycle misses a tool that appears and spreads between reviews. Discovery policies that alert on a new app turn the review into a standing process. Export a dated snapshot each quarter for the audit file. Sources: [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies). ## What should be left out of the inventory? Leave out the prompts themselves and any personal data about what an individual typed. The inventory records tools and metadata, not content. Where a product stores conversation data, restrict access: Nudge Security says conversation data is available by API for ingestion into customer platforms and incident response, so decide who may query it before enabling it. Keep usage counts at team level in widely shared reports. Individual names belong in the working record, where access can be limited, and not in a slide. Add a field for the tool's own claims, with the source and date, only when a decision depends on it. A vendor statement on training, hosting or retention should be stored with its URL so the reviewer can re-read it. That keeps the register honest when a vendor page later changes. Sources: [Nudge Security: AI conversation monitoring](https://www.nudgesecurity.com/features/ai-conversation-monitoring). ## Sources - [Cyberhaven: AI security](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 - [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai) (help.zscaler.com), read 2026-11-16 - [Aurascape: discover and monitor AI](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 - [Nudge Security: AI agent discovery](https://www.nudgesecurity.com/features/ai-agent-discovery) (nudgesecurity.com), read 2026-11-16 - [Auditing solutions in Microsoft Purview](https://learn.microsoft.com/en-us/purview/audit-solutions-overview) (learn.microsoft.com), read 2026-11-16 - [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix) (learn.microsoft.com), read 2026-11-16 - [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents) (learn.microsoft.com), read 2026-11-16 - [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies) (learn.microsoft.com), read 2026-11-16 - [Nudge Security: AI conversation monitoring](https://www.nudgesecurity.com/features/ai-conversation-monitoring) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Triaging What a Shadow AI Scan Finds](/guides/triaging-what-a-shadow-ai-scan-finds/) - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/) - [AI asset register (glossary)](/glossary/ai-asset-register/) - [All guides](/guides/) - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [The ranking](/ranking/) - [Glossary](/glossary/) --- # Shadow AI detection glossary URL: https://shadowaidetectiontools.com/glossary/ Updated: 2026-11-16 Twenty-eight shadow AI detection terms are defined here, including AI app catalog, OAuth grant, tenant restriction, TLS inspection, domain categorization and usage heatmap. Each entry gives a short definition and one sourced example read on 16 Nov 2026, with links to related terms. ## 28 terms - [AI app catalog](/glossary/ai-app-catalog/): An AI app catalog is a vendor-maintained database of known AI applications, with a record for each one that the detection tool matches traffic or sign-ins against. - [AI asset register](/glossary/ai-asset-register/): An AI asset register is a maintained list of the AI tools, agents, models and servers a company runs or allows, with an owner and a status for each. - [AI risk rating](/glossary/ai-risk-rating/): An AI risk rating is a score a tool assigns to each discovered AI app from factors such as data handling, vendor security posture and permissions requested. - [AI usage baseline](/glossary/ai-usage-baseline/): An AI usage baseline is a record of normal AI use in the organization, taken after a first scan, against which later change is judged. - [API-based discovery](/glossary/api-based-discovery/): API-based discovery finds AI tools by connecting to the APIs of SaaS apps and identity systems with read access. - [billing record discovery](/glossary/billing-record-discovery/): Billing record discovery finds software by reading expense reports, card statements or accounts-payable data for subscriptions. - [bring your own AI](/glossary/bring-your-own-ai/): Bring your own AI is the practice of employees using AI tools they chose and often pay for themselves to do company work. - [browser telemetry](/glossary/browser-telemetry/): Browser telemetry is data an extension reports about what a person does in the browser, such as which AI sites they open, which account they use and what they paste. - [CASB](/glossary/casb/): A cloud access security broker, or CASB, is a control point between users and cloud services that gives visibility into which services are used and enforces policy on them. - [discovery source](/glossary/discovery-source/): A discovery source is a data feed a detection tool reads to learn that an AI tool is in use, such as device inventory, browser activity, network traffic, sign-in logs, SaaS APIs or email. - [DNS-based detection](/glossary/dns-based-detection/): DNS-based detection finds AI tool use by reading or filtering the domain name lookups a device makes. - [domain categorization](/glossary/domain-categorization/): Domain categorization sorts web domains into categories, such as generative AI, so a policy can act on a whole class instead of one site. - [email receipt discovery](/glossary/email-receipt-discovery/): Email receipt discovery finds AI tools by scanning mailbox metadata for sign-up, verification and receipt messages. - [embedded AI feature](/glossary/embedded-ai-feature/): An embedded AI feature is an AI function switched on inside an application the company already approved, such as a summarizer in a document tool. - [just-in-time warning](/glossary/just-in-time-warning/): A just-in-time warning is a message shown at the moment an employee is about to use an unapproved AI tool or paste sensitive data into one. - [OAuth grant](/glossary/oauth-grant/): An OAuth grant is a permission an employee gives an app to act on a work account, such as reading mail or files in Google Workspace or Microsoft 365, without sharing a password. - [personal account usage](/glossary/personal-account-usage/): Personal account usage is an employee signing in to an AI service with a private account, not a company-managed one, while doing work. - [proxy log analysis](/glossary/proxy-log-analysis/): Proxy log analysis is reviewing the records a web proxy or firewall keeps of requests, to spot visits to AI services. - [SaaS discovery](/glossary/saas-discovery/): SaaS discovery is the process of finding which cloud applications employees use, whether or not IT bought them. - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/): Sanctioned AI is an AI tool the organization has reviewed and allowed; unsanctioned AI is any tool in use without that decision. - [shadow IT](/glossary/shadow-it/): Shadow IT is the use of technology, such as cloud apps and devices, that the IT or security team has not approved or does not know about. - [SSPM](/glossary/sspm/): SaaS security posture management, or SSPM, checks the configuration of connected SaaS apps against a baseline and flags risky settings such as open sharing or excess permissions. - [tenant restriction](/glossary/tenant-restriction/): A tenant restriction is a control that lets a company's sign-in traffic reach only approved tenants of a cloud service, so staff cannot sign in to a personal or outside account on the same service. - [TLS inspection](/glossary/tls-inspection/): TLS inspection is a gateway decrypting encrypted web traffic, reading it and encrypting it again, so a security tool can see paths, headers and request bodies. - [traffic fingerprinting](/glossary/traffic-fingerprinting/): Traffic fingerprinting identifies an AI service from patterns in its traffic, such as the shape of requests and responses, and not only from its domain name. - [training on inputs](/glossary/training-on-inputs/): Training on inputs means an AI vendor uses what customers type or upload to improve its models. - [usage heatmap](/glossary/usage-heatmap/): A usage heatmap is a view that shows AI use by team, tool or time, so a reviewer can see where use is concentrated. - [user nudge](/glossary/user-nudge/): A user nudge is a short message shown to an employee that steers them toward an approved choice without forbidding the tool. ## Related pages - [Guides](/guides/) - [FAQ](/faq/) - [The ranking](/ranking/) - [Vantage Point Matrix](/vantage-point-matrix/) - [Finding-to-Action Ladder](/finding-to-action-ladder/) --- # AI app catalog: definition and example URL: https://shadowaidetectiontools.com/glossary/ai-app-catalog/ Updated: 2026-11-16 An AI app catalog is a vendor-maintained database of known AI applications, with a record for each one that the detection tool matches traffic or sign-ins against. A tool can only name an AI app that is in its catalog or that it can classify on the fly. Catalog size and update speed therefore set how many unapproved tools a scan can label. ## Example Aurascape says it catalogs 20,000+ AI applications as of 24 June 2026, with identification, risk assessment and entitlement tier decoding for each. ## Sources - [Aurascape AI app discovery page](https://aurascape.ai/resources/blog/aurascape-ai-application-discovery/) (aurascape.ai), read 2026-11-16 ## Related pages - [AI risk rating](/glossary/ai-risk-rating/) - [SaaS discovery](/glossary/saas-discovery/) - [domain categorization](/glossary/domain-categorization/) - [All glossary terms](/glossary/) - [user nudge](/glossary/user-nudge/) - [AI asset register](/glossary/ai-asset-register/) - [Guides](/guides/) --- # AI asset register: definition and example URL: https://shadowaidetectiontools.com/glossary/ai-asset-register/ Updated: 2026-11-16 An AI asset register is a maintained list of the AI tools, agents, models and servers a company runs or allows, with an owner and a status for each. Discovery fills it, and a review keeps it current. Auditors ask for it as evidence that the organization knows what AI it has. ## Example Netskope says AI Command Center keeps an inventory of AI assets with risk scores, usage activity, threats and policy violations. ## Sources - [Netskope AI Command Center documentation](https://www.netskope.com/products/ai-command-center) (netskope.com), read 2026-11-16 ## Related pages - [AI risk rating](/glossary/ai-risk-rating/) - [usage heatmap](/glossary/usage-heatmap/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [How to Run a Shadow AI Audit, Step by Step](/guides/how-to-run-a-shadow-ai-audit/) - [Shadow AI Detection Evidence for Auditors](/guides/shadow-ai-detection-evidence-for-auditors/) - [What a Shadow AI Inventory Should Contain](/guides/what-a-shadow-ai-inventory-should-contain/) - [All glossary terms](/glossary/) - [AI app catalog](/glossary/ai-app-catalog/) - [Guides](/guides/) --- # AI risk rating: definition and example URL: https://shadowaidetectiontools.com/glossary/ai-risk-rating/ Updated: 2026-11-16 An AI risk rating is a score a tool assigns to each discovered AI app from factors such as data handling, vendor security posture and permissions requested. It lets a team sort hundreds of findings into a short list. The factors and weights differ by vendor, so ratings are not comparable across tools. ## Example On 30 September 2026 Nudge Security said adaptive risk management gives always-on app risk scores from 30 internal and external factors. ## Sources - [Nudge Security changelog](https://www.nudgesecurity.com/changelog) (nudgesecurity.com), read 2026-11-16 ## Related pages - [training on inputs](/glossary/training-on-inputs/) - [AI app catalog](/glossary/ai-app-catalog/) - [AI asset register](/glossary/ai-asset-register/) - [Triaging What a Shadow AI Scan Finds](/guides/triaging-what-a-shadow-ai-scan-finds/) - [All glossary terms](/glossary/) - [AI usage baseline](/glossary/ai-usage-baseline/) - [Guides](/guides/) --- # AI usage baseline: definition and example URL: https://shadowaidetectiontools.com/glossary/ai-usage-baseline/ Updated: 2026-11-16 An AI usage baseline is a record of normal AI use in the organization, taken after a first scan, against which later change is judged. A new tool, a spike in users or a move to personal accounts stands out against it. A baseline turns one-time discovery into ongoing monitoring. ## Example Aurascape publishes a sample AI security report that counts conversations, active users and active devices, ranks applications and users, and flags newly observed unsanctioned tools. ## Sources - [Aurascape sample AI security report](https://aurascape.ai/resources/sample-report/sample-ai-security-report/) (aurascape.ai), read 2026-11-16 ## Related pages - [usage heatmap](/glossary/usage-heatmap/) - [AI asset register](/glossary/ai-asset-register/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [All glossary terms](/glossary/) - [AI risk rating](/glossary/ai-risk-rating/) - [API-based discovery](/glossary/api-based-discovery/) - [Guides](/guides/) --- # API-based discovery: definition and example URL: https://shadowaidetectiontools.com/glossary/api-based-discovery/ Updated: 2026-11-16 API-based discovery finds AI tools by connecting to the APIs of SaaS apps and identity systems with read access. It needs no software on devices and can see AI already inside a connected tenant. It cannot see tools in apps that no one connected. ## Example On 23 July 2026 Cyberhaven said it reads ChatGPT Enterprise and Claude Enterprise through their Compliance APIs, covering conversations, uploaded files and custom GPT configurations. ## Sources - [Cyberhaven press release on ChatGPT Enterprise and Claude Enterprise](https://www.cyberhaven.com/press-releases/cyberhaven-extends-ai-security-chatgpt-enterprise-claude) (cyberhaven.com), read 2026-11-16 ## Related pages - [discovery source](/glossary/discovery-source/) - [OAuth grant](/glossary/oauth-grant/) - [embedded AI feature](/glossary/embedded-ai-feature/) - [All glossary terms](/glossary/) - [AI usage baseline](/glossary/ai-usage-baseline/) - [billing record discovery](/glossary/billing-record-discovery/) - [Guides](/guides/) --- # Billing record discovery: definition and example URL: https://shadowaidetectiontools.com/glossary/billing-record-discovery/ Updated: 2026-11-16 Billing record discovery finds software by reading expense reports, card statements or accounts-payable data for subscriptions. It catches paid AI tools even when no one signs in with a work identity, but it misses free tiers. The data comes from finance systems, which most security tools do not connect to. ## Example Zylo lists employee expense reports and credit card payments among the places where unapproved apps hide. ## Sources - [SaaS discovery, Zylo](https://www.zylo.com/blog/saas-discovery/) (zylo.com), read 2026-11-16 ## Related pages - [email receipt discovery](/glossary/email-receipt-discovery/) - [SaaS discovery](/glossary/saas-discovery/) - [shadow IT](/glossary/shadow-it/) - [All glossary terms](/glossary/) - [API-based discovery](/glossary/api-based-discovery/) - [bring your own AI](/glossary/bring-your-own-ai/) - [Guides](/guides/) --- # bring your own AI: definition and example URL: https://shadowaidetectiontools.com/glossary/bring-your-own-ai/ Updated: 2026-11-16 Bring your own AI is the practice of employees using AI tools they chose and often pay for themselves to do company work. It differs from a rollout the company runs, because the account, the terms and the data handling belong to the individual. Detection tools look for it by separating personal accounts from corporate ones. ## Example In a newswire release, Cyberhaven Labs reported an analysis of 3 million workers finding AI usage grew 485% year over year, with over 90% of use in personal shadow AI accounts (a 2024 report). ## Sources - [Cyberhaven Labs newswire release](https://www.prnewswire.com/news-releases/cyberhaven-report-surge-in-shadow-ai-accounts-poses-fresh-risks-to-corporate-data-302151221.html) (prnewswire.com), read 2026-11-16 ## Related pages - [personal account usage](/glossary/personal-account-usage/) - [tenant restriction](/glossary/tenant-restriction/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [All glossary terms](/glossary/) - [billing record discovery](/glossary/billing-record-discovery/) - [browser telemetry](/glossary/browser-telemetry/) - [Guides](/guides/) --- # Browser telemetry: definition and example URL: https://shadowaidetectiontools.com/glossary/browser-telemetry/ Updated: 2026-11-16 Browser telemetry is data an extension reports about what a person does in the browser, such as which AI sites they open, which account they use and what they paste. It sees activity that network tools cannot decode and works off the corporate network. It covers only browsers where the extension is installed. ## Example Push says it deploys as a lightweight extension in existing browsers and sees browser-layer activity that network and endpoint tools lack. ## Sources - [Push Security shadow AI guide](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps) (pushsecurity.com), read 2026-11-16 ## Related pages - [personal account usage](/glossary/personal-account-usage/) - [just-in-time warning](/glossary/just-in-time-warning/) - [discovery source](/glossary/discovery-source/) - [All glossary terms](/glossary/) - [bring your own AI](/glossary/bring-your-own-ai/) - [CASB](/glossary/casb/) - [Guides](/guides/) --- # CASB: definition and example URL: https://shadowaidetectiontools.com/glossary/casb/ Updated: 2026-11-16 A cloud access security broker, or CASB, is a control point between users and cloud services that gives visibility into which services are used and enforces policy on them. Shadow IT discovery is one of its basic functions, and many shadow AI tools grew from it. It can sit inline as a proxy or connect through APIs. ## Example Microsoft says Defender for Cloud Apps provides fundamental CASB functionality, such as shadow IT discovery, visibility into cloud app usage and protection against app-based threats. ## Sources - [What is Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/what-is-defender-for-cloud-apps) (learn.microsoft.com), read 2026-11-16 ## Related pages - [SSPM](/glossary/sspm/) - [SaaS discovery](/glossary/saas-discovery/) - [shadow IT](/glossary/shadow-it/) - [Shadow AI Detection in Microsoft 365 and Google Workspace](/guides/shadow-ai-detection-in-microsoft-365-and-google-workspace/) - [All glossary terms](/glossary/) - [browser telemetry](/glossary/browser-telemetry/) - [discovery source](/glossary/discovery-source/) - [Guides](/guides/) --- # Discovery source: definition and example URL: https://shadowaidetectiontools.com/glossary/discovery-source/ Updated: 2026-11-16 A discovery source is a data feed a detection tool reads to learn that an AI tool is in use, such as device inventory, browser activity, network traffic, sign-in logs, SaaS APIs or email. Each source sees a different slice, so the mix of sources decides what a first inventory contains. Vendors usually name their sources on a product page. ## Example Reco lists its shadow AI detection methods as browser behavior, OAuth connections, email metadata, network traffic analysis and browser extension detection. ## Sources - [Reco shadow AI discovery page](https://www.reco.ai/use-cases/shadow-ai-discovery) (reco.ai), read 2026-11-16 ## Related pages - [API-based discovery](/glossary/api-based-discovery/) - [browser telemetry](/glossary/browser-telemetry/) - [email receipt discovery](/glossary/email-receipt-discovery/) - [All glossary terms](/glossary/) - [CASB](/glossary/casb/) - [DNS-based detection](/glossary/dns-based-detection/) - [Guides](/guides/) --- # DNS-based detection: definition and example URL: https://shadowaidetectiontools.com/glossary/dns-based-detection/ Updated: 2026-11-16 DNS-based detection finds AI tool use by reading or filtering the domain name lookups a device makes. It works across protocols and applications, since every connection starts with a lookup, but it sees only the domain and not the account or the content. It suits a first scan of which AI domains are being reached. ## Example Cloudflare's documentation says DNS policies let you control which websites and services users can reach by inspecting DNS queries, and that they act at the lookup stage across all protocols and applications. ## Sources - [DNS policies, Cloudflare One docs](https://developers.cloudflare.com/cloudflare-one/traffic-policies/dns-policies/) (developers.cloudflare.com), read 2026-11-16 ## Related pages - [domain categorization](/glossary/domain-categorization/) - [proxy log analysis](/glossary/proxy-log-analysis/) - [TLS inspection](/glossary/tls-inspection/) - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [All glossary terms](/glossary/) - [discovery source](/glossary/discovery-source/) - [Guides](/guides/) --- # Domain categorization: definition and example URL: https://shadowaidetectiontools.com/glossary/domain-categorization/ Updated: 2026-11-16 Domain categorization sorts web domains into categories, such as generative AI, so a policy can act on a whole class instead of one site. Coverage depends on how fast new AI domains are added. It is the simplest way to block or flag a long tail of AI tools. ## Example On 2 June 2026 Push launched file, clipboard, app and domain controls for unsanctioned AI tools, with an 89-category domain framework. ## Sources - [Push Security AI visibility and control announcement](https://pushsecurity.com/news/push-ai-visibility-and-control-capabilities) (pushsecurity.com), read 2026-11-16 ## Related pages - [DNS-based detection](/glossary/dns-based-detection/) - [AI app catalog](/glossary/ai-app-catalog/) - [traffic fingerprinting](/glossary/traffic-fingerprinting/) - [All glossary terms](/glossary/) - [email receipt discovery](/glossary/email-receipt-discovery/) - [Guides](/guides/) --- # Email receipt discovery: definition and example URL: https://shadowaidetectiontools.com/glossary/email-receipt-discovery/ Updated: 2026-11-16 Email receipt discovery finds AI tools by scanning mailbox metadata for sign-up, verification and receipt messages. A tool a person signed up for with a work address leaves such a trail even if it never touches the network. It needs read access to mail, which raises a privacy question. ## Example Nudge Security says discovery starts from email analysis through a Microsoft 365 or Google Workspace integration, then adds a browser extension and SaaS API integrations. ## Sources - [Nudge Security AI security page](https://www.nudgesecurity.com/use-cases/ai-security-3) (nudgesecurity.com), read 2026-11-16 ## Related pages - [billing record discovery](/glossary/billing-record-discovery/) - [discovery source](/glossary/discovery-source/) - [API-based discovery](/glossary/api-based-discovery/) - [All glossary terms](/glossary/) - [domain categorization](/glossary/domain-categorization/) - [embedded AI feature](/glossary/embedded-ai-feature/) - [Guides](/guides/) --- # embedded AI feature: definition and example URL: https://shadowaidetectiontools.com/glossary/embedded-ai-feature/ Updated: 2026-11-16 An embedded AI feature is an AI function switched on inside an application the company already approved, such as a summarizer in a document tool. The web address stays the same, so traffic-based tools see only the approved destination. Finding these features usually takes an app-level or API-level view. ## Example Aurascape says it finds AI embedded inside trusted SaaS apps and websites, a case where traditional tools only see the destination. ## Sources - [Aurascape discover and monitor AI page](https://aurascape.ai/see-and-control-ai-security-platform/) (aurascape.ai), read 2026-11-16 ## Related pages - [SaaS discovery](/glossary/saas-discovery/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [API-based discovery](/glossary/api-based-discovery/) - [Embedded AI Features Switched On in Approved SaaS](/guides/embedded-ai-features-switched-on-in-approved-saas/) - [All glossary terms](/glossary/) - [email receipt discovery](/glossary/email-receipt-discovery/) - [just-in-time warning](/glossary/just-in-time-warning/) - [Guides](/guides/) --- # Just-in-time warning: definition and example URL: https://shadowaidetectiontools.com/glossary/just-in-time-warning/ Updated: 2026-11-16 A just-in-time warning is a message shown at the moment an employee is about to use an unapproved AI tool or paste sensitive data into one. Because it appears in the flow of work, it can explain the rule and offer an approved option on the spot. It needs a sensor on the device or browser that can interrupt the action. ## Example Harmonic's pricing page says its Guide feature warns, nudges or blocks inline in under 200 milliseconds. ## Sources - [Harmonic Security pricing page](https://www.harmonic.security/pricing) (harmonic.security), read 2026-11-16 ## Related pages - [user nudge](/glossary/user-nudge/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [browser telemetry](/glossary/browser-telemetry/) - [All glossary terms](/glossary/) - [embedded AI feature](/glossary/embedded-ai-feature/) - [OAuth grant](/glossary/oauth-grant/) - [Guides](/guides/) --- # OAuth grant: definition and example URL: https://shadowaidetectiontools.com/glossary/oauth-grant/ Updated: 2026-11-16 An OAuth grant is a permission an employee gives an app to act on a work account, such as reading mail or files in Google Workspace or Microsoft 365, without sharing a password. AI assistants and agents often connect this way, and the grant stays active until someone removes it. Listing grants is one way to find AI tools that never show up in browser traffic. ## Example Nudge Security says its browser extension detects OAuth grants that power MCP server integrations, which covers remote MCP servers connected to AI tools. ## Sources - [Nudge Security shadow AI MCP page](https://www.nudgesecurity.com/post/how-to-monitor-remote-mcp-server-connections-with-nudge-security) (nudgesecurity.com), read 2026-11-16 ## Related pages - [API-based discovery](/glossary/api-based-discovery/) - [discovery source](/glossary/discovery-source/) - [embedded AI feature](/glossary/embedded-ai-feature/) - [Shadow AI Discovery From Identity Provider Logs](/guides/shadow-ai-discovery-from-identity-provider-logs/) - [All glossary terms](/glossary/) - [just-in-time warning](/glossary/just-in-time-warning/) - [personal account usage](/glossary/personal-account-usage/) - [Guides](/guides/) --- # Personal account usage: definition and example URL: https://shadowaidetectiontools.com/glossary/personal-account-usage/ Updated: 2026-11-16 Personal account usage is an employee signing in to an AI service with a private account, not a company-managed one, while doing work. The company has no admin controls, no audit log and usually no data-handling terms for that account. Detecting it needs a view of which account is logged in, which a browser extension can provide. ## Example Bay says it identifies personal versus corporate AI accounts including in personal Chrome profiles; no full browser list was found. ## Sources - [Bay browser platform page](https://www.bay.io/platform/browser?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=term) (bay.io), read 2026-11-16 ## Related pages - [bring your own AI](/glossary/bring-your-own-ai/) - [tenant restriction](/glossary/tenant-restriction/) - [browser telemetry](/glossary/browser-telemetry/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [All glossary terms](/glossary/) - [OAuth grant](/glossary/oauth-grant/) - [proxy log analysis](/glossary/proxy-log-analysis/) - [Guides](/guides/) --- # Proxy log analysis: definition and example URL: https://shadowaidetectiontools.com/glossary/proxy-log-analysis/ Updated: 2026-11-16 Proxy log analysis is reviewing the records a web proxy or firewall keeps of requests, to spot visits to AI services. It covers every device behind the gateway without installing software, but only while the device is on that path. Logs are matched to an app catalog to turn domains into named AI tools. ## Example Microsoft documents three cloud discovery feeds: the Defender for Endpoint integration on Windows 10 and 11 devices, a log collector on firewalls and proxies, and proxy integrations such as Zscaler. ## Sources - [Microsoft Defender for Cloud Apps shadow IT tutorial](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it) (learn.microsoft.com), read 2026-11-16 ## Related pages - [DNS-based detection](/glossary/dns-based-detection/) - [TLS inspection](/glossary/tls-inspection/) - [AI app catalog](/glossary/ai-app-catalog/) - [Reading Proxy and DNS Logs for AI Apps](/guides/reading-proxy-and-dns-logs-for-ai-apps/) - [All glossary terms](/glossary/) - [personal account usage](/glossary/personal-account-usage/) - [SaaS discovery](/glossary/saas-discovery/) - [Guides](/guides/) --- # SaaS discovery: definition and example URL: https://shadowaidetectiontools.com/glossary/saas-discovery/ Updated: 2026-11-16 SaaS discovery is the process of finding which cloud applications employees use, whether or not IT bought them. For AI it extends to chat assistants, AI features inside approved apps and agents connected by OAuth. It usually starts from traffic, identity or email data matched against an app catalog. ## Example Microsoft says Defender for Cloud Apps discovers cloud and SaaS apps from network traffic against an app catalog and ranks each app's risk. ## Sources - [Microsoft Defender for Cloud Apps shadow IT tutorial](https://learn.microsoft.com/en-us/defender-cloud-apps/what-is-defender-for-cloud-apps) (learn.microsoft.com), read 2026-11-16 ## Related pages - [shadow IT](/glossary/shadow-it/) - [AI app catalog](/glossary/ai-app-catalog/) - [CASB](/glossary/casb/) - [All glossary terms](/glossary/) - [proxy log analysis](/glossary/proxy-log-analysis/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [Guides](/guides/) --- # sanctioned vs unsanctioned AI: definition and example URL: https://shadowaidetectiontools.com/glossary/sanctioned-vs-unsanctioned-ai/ Updated: 2026-11-16 Sanctioned AI is an AI tool the organization has reviewed and allowed; unsanctioned AI is any tool in use without that decision. Some tools add middle states such as tolerated or restricted for tools that are allowed with limits. The label drives which action a detection tool takes on a finding. ## Example Cyberhaven says it categorizes AI applications as Sanctioned, Unsanctioned, Tolerated or Restricted and shows adoption trends. ## Sources - [Cyberhaven AI security product page](https://www.cyberhaven.com/product/ai-security) (cyberhaven.com), read 2026-11-16 ## Related pages - [AI asset register](/glossary/ai-asset-register/) - [just-in-time warning](/glossary/just-in-time-warning/) - [bring your own AI](/glossary/bring-your-own-ai/) - [All glossary terms](/glossary/) - [SaaS discovery](/glossary/saas-discovery/) - [shadow IT](/glossary/shadow-it/) - [Guides](/guides/) --- # shadow IT: definition and example URL: https://shadowaidetectiontools.com/glossary/shadow-it/ Updated: 2026-11-16 Shadow IT is the use of technology, such as cloud apps and devices, that the IT or security team has not approved or does not know about. Shadow AI is a newer case of it, and many detection methods were built for shadow IT first. The older methods count apps; the newer ones have to read what is shared with them. ## Example Microsoft's Defender for Cloud Apps tutorial says its Defender for Endpoint integration collects data on cloud traffic across Windows 10 and Windows 11 devices, on and off the network. ## Sources - [Discover and manage Shadow IT, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it) (learn.microsoft.com), read 2026-11-16 ## Related pages - [SaaS discovery](/glossary/saas-discovery/) - [CASB](/glossary/casb/) - [proxy log analysis](/glossary/proxy-log-analysis/) - [Shadow AI vs Shadow IT: What Changes for Detection](/guides/shadow-ai-vs-shadow-it-what-changes-for-detection/) - [All glossary terms](/glossary/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [SSPM](/glossary/sspm/) - [Guides](/guides/) --- # SSPM: definition and example URL: https://shadowaidetectiontools.com/glossary/sspm/ Updated: 2026-11-16 SaaS security posture management, or SSPM, checks the configuration of connected SaaS apps against a baseline and flags risky settings such as open sharing or excess permissions. It reads app settings through APIs and does not watch user traffic. In shadow AI work it shows what connected AI apps are allowed to do inside a tenant. ## Example Help Net Security reported Grip's SSPM launch on 14 February 2025, describing a posture management product built on visibility and automation. ## Sources - [Help Net Security report on Grip SSPM](https://www.helpnetsecurity.com/2025/02/14/grip-security-unveils-sspm-solution-to-strengthen-saas-security-posture/) (helpnetsecurity.com), read 2026-11-16 ## Related pages - [CASB](/glossary/casb/) - [OAuth grant](/glossary/oauth-grant/) - [API-based discovery](/glossary/api-based-discovery/) - [All glossary terms](/glossary/) - [shadow IT](/glossary/shadow-it/) - [tenant restriction](/glossary/tenant-restriction/) - [Guides](/guides/) --- # Tenant restriction: definition and example URL: https://shadowaidetectiontools.com/glossary/tenant-restriction/ Updated: 2026-11-16 A tenant restriction is a control that lets a company's sign-in traffic reach only approved tenants of a cloud service, so staff cannot sign in to a personal or outside account on the same service. It is enforced at the network or device through a header or policy. It limits personal account use without blocking the service itself. ## Example Microsoft says organizations can use tenant restrictions to prevent a user from using some or all external apps while signed in with an external account on their network or devices. ## Sources - [Configure Tenant Restrictions, Microsoft Entra](https://learn.microsoft.com/en-us/entra/external-id/tenant-restrictions-v2) (learn.microsoft.com), read 2026-11-16 ## Related pages - [personal account usage](/glossary/personal-account-usage/) - [bring your own AI](/glossary/bring-your-own-ai/) - [proxy log analysis](/glossary/proxy-log-analysis/) - [Personal vs Corporate AI Accounts: Telling Them Apart](/guides/personal-versus-corporate-ai-accounts/) - [All glossary terms](/glossary/) - [SSPM](/glossary/sspm/) - [TLS inspection](/glossary/tls-inspection/) - [Guides](/guides/) --- # TLS inspection: definition and example URL: https://shadowaidetectiontools.com/glossary/tls-inspection/ Updated: 2026-11-16 TLS inspection is a gateway decrypting encrypted web traffic, reading it and encrypting it again, so a security tool can see paths, headers and request bodies. For AI detection it is what lets a proxy see what is typed into a chat tool and not only that the site was visited. It needs a certificate installed on devices and adds a privacy decision. ## Example Cloudflare's documentation says that without TLS decryption the full URL, headers and request body inside HTTPS are not visible to its Gateway. ## Sources - [TLS decryption, Cloudflare One docs](https://developers.cloudflare.com/cloudflare-one/traffic-policies/http-policies/tls-decryption/) (developers.cloudflare.com), read 2026-11-16 ## Related pages - [proxy log analysis](/glossary/proxy-log-analysis/) - [DNS-based detection](/glossary/dns-based-detection/) - [traffic fingerprinting](/glossary/traffic-fingerprinting/) - [All glossary terms](/glossary/) - [tenant restriction](/glossary/tenant-restriction/) - [Guides](/guides/) --- # Traffic fingerprinting: definition and example URL: https://shadowaidetectiontools.com/glossary/traffic-fingerprinting/ Updated: 2026-11-16 Traffic fingerprinting identifies an AI service from patterns in its traffic, such as the shape of requests and responses, and not only from its domain name. It helps when a service uses shared hosting or an address that a domain list misses. It is used alongside a catalog and, where allowed, decryption. ## Example Aurascape says its discovery works by inspecting and categorizing AI traffic in motion, reading queries and responses. ## Sources - [Aurascape discover and monitor AI page](https://aurascape.ai/discover-and-monitor-ai/) (aurascape.ai), read 2026-11-16 ## Related pages - [TLS inspection](/glossary/tls-inspection/) - [domain categorization](/glossary/domain-categorization/) - [AI app catalog](/glossary/ai-app-catalog/) - [Shadow AI False Positives and Blind Spots](/guides/shadow-ai-false-positives-and-blind-spots/) - [All glossary terms](/glossary/) - [training on inputs](/glossary/training-on-inputs/) - [Guides](/guides/) --- # Training on inputs: definition and example URL: https://shadowaidetectiontools.com/glossary/training-on-inputs/ Updated: 2026-11-16 Training on inputs means an AI vendor uses what customers type or upload to improve its models. Whether it does depends on the plan and the contract, and consumer plans often differ from enterprise ones. It is a core factor in judging a discovered AI tool, because pasted data may not stay private. ## Example Harmonic says each tool has a profile of who uses it, how often, and whether the vendor trains on user inputs. ## Sources - [Harmonic Security product page](https://www.harmonic.security/products/explore) (harmonic.security), read 2026-11-16 ## Related pages - [AI risk rating](/glossary/ai-risk-rating/) - [personal account usage](/glossary/personal-account-usage/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [All glossary terms](/glossary/) - [traffic fingerprinting](/glossary/traffic-fingerprinting/) - [usage heatmap](/glossary/usage-heatmap/) - [Guides](/guides/) --- # Usage heatmap: definition and example URL: https://shadowaidetectiontools.com/glossary/usage-heatmap/ Updated: 2026-11-16 A usage heatmap is a view that shows AI use by team, tool or time, so a reviewer can see where use is concentrated. It turns an inventory list into a picture of who relies on which tool. Security teams use it to decide which findings to look at first. ## Example Zscaler documents an overview showing per-app policy actions (allowed, blocked or isolated), usage by department and top users, with drill-down for incident response. ## Sources - [Zscaler Zero Trust Exchange AI security documentation](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-overview) (help.zscaler.com), read 2026-11-16 ## Related pages - [AI usage baseline](/glossary/ai-usage-baseline/) - [AI asset register](/glossary/ai-asset-register/) - [AI risk rating](/glossary/ai-risk-rating/) - [Shadow AI Metrics for the Board](/guides/shadow-ai-metrics-for-the-board/) - [All glossary terms](/glossary/) - [training on inputs](/glossary/training-on-inputs/) - [user nudge](/glossary/user-nudge/) - [Guides](/guides/) --- # User nudge: definition and example URL: https://shadowaidetectiontools.com/glossary/user-nudge/ Updated: 2026-11-16 A user nudge is a short message shown to an employee that steers them toward an approved choice without forbidding the tool. It relies on persuasion and a clear alternative. Nudges suit early rollouts where blocking would send staff to harder-to-see workarounds. ## Example Nudge Security says its browser extension nudges employees away from unapproved apps toward sanctioned tools and delivers the AI acceptable use policy. ## Sources - [Nudge Security AI security page](https://www.nudgesecurity.com/features/browser-extension) (nudgesecurity.com), read 2026-11-16 ## Related pages - [just-in-time warning](/glossary/just-in-time-warning/) - [sanctioned vs unsanctioned AI](/glossary/sanctioned-vs-unsanctioned-ai/) - [personal account usage](/glossary/personal-account-usage/) - [All glossary terms](/glossary/) - [usage heatmap](/glossary/usage-heatmap/) - [AI app catalog](/glossary/ai-app-catalog/) - [Guides](/guides/) --- # Shadow AI detection news URL: https://shadowaidetectiontools.com/news/ Updated: 2026-11-16 Shadow AI detection news is collected from primary sources: vendor releases, funding and acquisition announcements, studies on unapproved AI use with the sample size stated, and changes to AI app catalogs and OAuth consent controls. Each item names its source and the date it was read. ## 12 items - 2026-11-16: [Nudge Security adds OpenAI Agents connected app](/news/nudge-security-adds-openai-agents-connected-app/) - 2026-11-09: [Nudge Security adds regional data checks for AI chats](/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/) - 2026-11-05: [Cyberhaven launches Agentic Data Security module](/news/cyberhaven-launches-agentic-data-security/) - 2026-11-02: [Nudge Security discovers AI agents across 23 platforms](/news/nudge-security-ai-agent-discovery-across-23-platforms/) - 2026-10-26: [Sophos launches AI Defense to find and control shadow AI](/news/sophos-launches-ai-defense-for-shadow-ai/) - 2026-10-21: [O'Reilly survey links shadow AI to policy strength](/news/oreilly-security-survey-shadow-ai-and-policy-strength/) - 2026-10-16: [Zscaler adds MCP server policies to AI Asset Management](/news/zscaler-ai-asset-management-adds-mcp-server-policies/) - 2026-10-14: [Harmonic Security raises $50 million Series B](/news/harmonic-security-raises-50-million-series-b/) - 2026-10-07: [Aurascape adds masking and model-level policy](/news/aurascape-september-release-masking-and-model-level-policy/) - 2026-10-05: [Zscaler describes Endpoint AI Security for shadow AI](/news/zscaler-endpoint-ai-security-discovers-ai-on-managed-devices/) - 2026-09-29: [Reco raises $55 million as it widens from SaaS to agents](/news/reco-raises-55-million-agent-security/) - 2026-08-19: [Zscaler adds AI Security policies for cloud agent platforms](/news/zscaler-adds-ai-security-policies-for-bedrock-foundry-copilot-studio/) Subscribe with the [RSS feed](/feed.xml). ## Related pages - [Guides](/guides/) - [The ranking](/ranking/) - [Vendors](/vendors/) - [Send a correction or tip](/corrections/) --- # Nudge Security adds OpenAI Agents connected app URL: https://shadowaidetectiontools.com/news/nudge-security-adds-openai-agents-connected-app/ Updated: 2026-11-16 Nudge Security's changelog entry of 16 November 2026 says it now discovers AI agents that teams build on the OpenAI Agents API. Admins enable a dedicated OpenAI Agents connected app, and each agent appears in the AI agents inventory under Identities with its creator, components and risk insights. 2026-11-16 The changelog entry is dated 16 November 2026 and is labeled a feature update under AI security. To start, an administrator enables the dedicated OpenAI Agents connected app in Nudge Security. Each discovered agent flows into the AI agents inventory under Identities, where the entry says admins see its creator, components and risk insights. This extends the API-based discovery list the company published on 2 November. A connected app is an approved integration, so the discovery depends on an administrator granting access; agents built outside that connection would need another discovery source such as the browser extension. The entry gives no price. ## Sources - [Nudge Security changelog](https://www.nudgesecurity.com/changelog) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Nudge Security profile](/vendors/nudge-security/) - [All news](/news/) - [Zscaler adds AI Security policies for cloud agent platforms](/news/zscaler-adds-ai-security-policies-for-bedrock-foundry-copilot-studio/) - [Nudge Security adds regional data checks for AI chats](/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/) - [The ranking](/ranking/) --- # Nudge Security adds regional data checks for AI chats URL: https://shadowaidetectiontools.com/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/ Updated: 2026-11-09 Nudge Security's changelog of 9 November 2026 says AI conversation monitoring now detects region-specific sensitive data, such as national ID, tax and health insurance numbers, on top of PII, secrets and financial information. Admins choose which regions to turn on, and the entry says the data can also be stopped from being shared. 2026-11-09 The entry lists regions admins can enable: North America (Canadian SIN, Mexican CURP and NSS), South America (Brazil CPF and CNPJ, Argentina CUIT and CUIL, Chile RUN and RUT), Europe (national IDs and tax numbers from Germany, Italy, Poland and Spain, plus UK NHS numbers), Asia Pacific (India Aadhaar, Singapore NRIC, Japan My Number, Australia TFN and Medicare) and Israel and South Africa national IDs. The monitoring is part of the Nudge Security browser extension, so it sees conversations in browsers where the extension is installed. A related entry on 3 November added detection of cryptocurrency wallet addresses shared in AI conversations. ## Sources - [Nudge Security changelog](https://www.nudgesecurity.com/changelog) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Nudge Security profile](/vendors/nudge-security/) - [All news](/news/) - [Nudge Security adds OpenAI Agents connected app](/news/nudge-security-adds-openai-agents-connected-app/) - [Cyberhaven launches Agentic Data Security module](/news/cyberhaven-launches-agentic-data-security/) - [The ranking](/ranking/) --- # Cyberhaven launches Agentic Data Security module URL: https://shadowaidetectiontools.com/news/cyberhaven-launches-agentic-data-security/ Updated: 2026-11-05 Cyberhaven introduced Agentic Data Security on 5 November 2026 and says it is available for purchase. Cyberhaven says it discovers AI tools and agents across endpoints, any cloud or browser, including unsanctioned apps, scores each for risk, captures prompts and tool calls, and applies existing data policies to agent actions. 2026-11-05 Cyberhaven's release of 5 November 2026 describes Agentic Data Security as the latest module in Cyberhaven Flow, its data loss prevention, insider risk and data security posture platform. It lists four controls: find every agent, see what it does, know who sent what, and stop data before it leaves. The release cites Cyberhaven Labs, saying endpoint-based agentic AI app adoption doubled year over year, reaching 60% in May 2026. The release does not state the sample behind that figure, so treat it as a vendor measure of its own telemetry. It quotes Dan Walsh, CISO of Datavant, and Lawrence Pingree of SACR. No price is stated. ## Sources - [Cyberhaven delivers Agentic Data Security](https://www.cyberhaven.com/press-releases/cyberhaven-delivers-agentic-data-security) (cyberhaven.com), read 2026-11-16 ## Related pages - [Cyberhaven profile](/vendors/cyberhaven/) - [All news](/news/) - [Nudge Security adds regional data checks for AI chats](/news/nudge-security-detects-region-specific-sensitive-data-in-ai-conversations/) - [Nudge Security discovers AI agents across 23 platforms](/news/nudge-security-ai-agent-discovery-across-23-platforms/) - [The ranking](/ranking/) --- # Nudge Security discovers AI agents across 23 platforms URL: https://shadowaidetectiontools.com/news/nudge-security-ai-agent-discovery-across-23-platforms/ Updated: 2026-11-02 Nudge Security's changelog entry of 2 November 2026 says AI agent discovery across 23 agentic platforms is available to all customers. Twelve platforms are found through API connections, including ChatGPT, Gemini, Microsoft Copilot Studio and Salesforce Agentforce, and eleven through the browser extension, including Cursor automations and Zapier Agents. 2026-11-02 Nudge Security's changelog says on 2 November 2026 that it discovers AI agents across 23 agentic platforms through API connections and the Nudge Security browser extension, and that the capability is available to all customers. API-based discovery covers Abacus.AI, ChatGPT, Claude managed agents, Gemini, LangSmith Fleet, Microsoft Copilot Studio, n8n, Notion AI agents, Salesforce Agentforce, ServiceNow, Tines and Workato. Browser-based discovery covers Amazon Quick, Atlassian Rovo, ChatGPT workspace agents, Cursor automations, Datadog, HubSpot, HyperAgent, OpenAI Agents Workflows, Retool, Zapier Agents and Zoom AI Workflows. The split matters for buyers: API discovery needs the app's administrator to connect it, while browser-based discovery sees agents on platforms without public APIs but only in browsers where the extension is installed. ## Sources - [Nudge Security changelog](https://www.nudgesecurity.com/changelog) (nudgesecurity.com), read 2026-11-16 ## Related pages - [Nudge Security profile](/vendors/nudge-security/) - [All news](/news/) - [Cyberhaven launches Agentic Data Security module](/news/cyberhaven-launches-agentic-data-security/) - [Sophos launches AI Defense to find and control shadow AI](/news/sophos-launches-ai-defense-for-shadow-ai/) - [The ranking](/ranking/) --- # Sophos launches AI Defense to find and control shadow AI URL: https://shadowaidetectiontools.com/news/sophos-launches-ai-defense-for-shadow-ai/ Updated: 2026-10-26 Sophos announced general availability of AI Defense, Campus Technology reported on 26 October 2026. Sophos says it discovers AI tools, agents and workflows across users and devices, including shadow AI, using data from devices, networks, identity systems and browsers, and can inspect prompts and uploads to help prevent sensitive data exposure. 2026-10-26 Campus Technology reported on 26 October 2026 that Sophos made AI Defense generally available. The offering runs through Sophos Fusion and uses information from devices, networks, identity systems and browsers. Sophos says it can inspect prompts, uploads and other AI interactions. Sophos' product page, read on 16 Nov 2026, lists four jobs: see AI activity (discover AI tools, agents and workflows, including shadow AI), control AI use, reduce AI risk by assessing identity, permissions, behavior, data exposure and location, and stop AI threats through integrated detection and response workflows. The page cites a Gartner figure that 69% of organizations suspect or have evidence of prohibited public AI use, and a Cloud Security Alliance figure that 82% discovered AI tools or workflows they did not know existed. The page carries no launch date and no price, so the 26 October date comes from the trade report. ## Sources - [Sophos AI Defense](https://www.sophos.com/en-us/products/ai-defense) (sophos.com), read 2026-11-16 - [Sophos Launches Tool to Defend Against Shadow AI, Campus Technology](https://campustechnology.com/articles/2026/10/26/sophos-launches-tool-to-defend-against-shadow-ai.aspx) (campustechnology.com), read 2026-11-16 ## Related pages - [All news](/news/) - [Nudge Security discovers AI agents across 23 platforms](/news/nudge-security-ai-agent-discovery-across-23-platforms/) - [O'Reilly survey links shadow AI to policy strength](/news/oreilly-security-survey-shadow-ai-and-policy-strength/) - [The ranking](/ranking/) --- # O'Reilly survey links shadow AI to policy strength URL: https://shadowaidetectiontools.com/news/oreilly-security-survey-shadow-ai-and-policy-strength/ Updated: 2026-10-21 O'Reilly published its AI security survey on 21 October 2026: 351 respondents completed it, 209 of them involved in security. Of those, 36% said teammates occasionally use shadow AI and 23% said it is widespread. Widespread use was reported by 17% with a comprehensive policy and 35% with a basic one. 2026-10-21 O'Reilly's Radar article, dated 21 October 2026, reports a survey in which 351 respondents completed the full questionnaire and 209 of them self-identified as involved in security. Help Net Security's coverage on 26 October describes the same 209 security professionals, surveyed in September. On shadow AI, defined in the survey as AI use outside any definitive policy, 36% thought teammates occasionally use it, 23% thought the practice is widespread, 24% said it is rare, and 16% were unsure or lacked monitoring tools. Among respondents with a comprehensive AI policy, 17% said shadow AI is widespread, against 35% with a basic policy. O'Reilly's author notes shadow AI is probably underreported. The survey measures what respondents believe about teammates, not what a tool detected, so it is a reference for scale and not a substitute for a scan of your own environment. ## Sources - [The State of AI Security: Threats, Tools, and Skills Gaps, O'Reilly Radar](https://www.oreilly.com/radar/the-state-of-ai-security-threats-tools-and-skills-gaps/) (oreilly.com), read 2026-11-16 - [Help Net Security coverage of the O'Reilly AI security survey](https://www.helpnetsecurity.com/2026/10/26/ai-tools-security-worry-survey/) (helpnetsecurity.com), read 2026-11-16 ## Related pages - [All news](/news/) - [Sophos launches AI Defense to find and control shadow AI](/news/sophos-launches-ai-defense-for-shadow-ai/) - [Zscaler adds MCP server policies to AI Asset Management](/news/zscaler-ai-asset-management-adds-mcp-server-policies/) - [The ranking](/ranking/) --- # Zscaler adds MCP server policies to AI Asset Management URL: https://shadowaidetectiontools.com/news/zscaler-ai-asset-management-adds-mcp-server-policies/ Updated: 2026-10-16 Zscaler's AI Asset Management release notes dated 16 October 2026 add MCP server policies. Titles listed are an MCP server with critical risk level, an MCP server that performs file deletion operations, and an MCP server that exposes code execution tools. The same entry adds checks for Microsoft Foundry and Azure OpenAI guardrail settings. 2026-10-16 The release notes entry for 16 October 2026 lists new AI Security policies under several asset types. For MCP servers the titles are: MCP server with critical risk level, MCP server performs file deletion operations, and MCP server exposes code execution tools. The same entry adds policies for AI models (a model whose country of origin is China, and a model whose supply chain risk score is above 80) and for Microsoft Foundry or Azure OpenAI guardrails whose hate, violence, sexual and self-harm filters are not set to the highest blocking level. These are posture policies on assets the product has already inventoried. The notes do not say how MCP servers on laptops are found, and they give no availability tiers or prices. ## Sources - [Release Upgrade Summary (2026), Zscaler AI Asset Management](https://help.zscaler.com/ai-asset-mgmt/release-upgrade-summary-2026) (help.zscaler.com), read 2026-11-16 ## Related pages - [Zscaler profile](/vendors/zscaler/) - [All news](/news/) - [O'Reilly survey links shadow AI to policy strength](/news/oreilly-security-survey-shadow-ai-and-policy-strength/) - [Harmonic Security raises $50 million Series B](/news/harmonic-security-raises-50-million-series-b/) - [The ranking](/ranking/) --- # Harmonic Security raises $50 million Series B URL: https://shadowaidetectiontools.com/news/harmonic-security-raises-50-million-series-b/ Updated: 2026-10-14 Harmonic Security announced a $50 million Series B on 14 October 2026, SecurityWeek reported. Crosspoint Capital Partners led the round, and total funding reaches $76 million. Harmonic says its platform covers browser-based AI, desktop AI tools and agentic workflows, and uses small language models to read each AI interaction in context. 2026-10-14 SecurityWeek reported on 14 October 2026 that Harmonic Security raised $50 million in a Series B round, bringing the total raised to $76 million. Crosspoint Capital Partners led, with support from existing investors. The money is for engineering and go-to-market hiring in the US and UK. Harmonic was founded in 2023 by Alastair Paterson (CEO) and Bryan Woolgar-O'Neil (CTO), who previously founded Digital Shadows. The platform covers browser-based AI, desktop AI tools and agentic workflows, and exposes risks such as personal-account use and sensitive data exposure. Harmonic's Usage Intelligence layer, which shows the AI use cases that create value, runs on the same small-language-model engine. Unite.AI separately reported Harmonic's statement that it protects millions of employees at more than 100 companies. A Harmonic-hosted release page for the round was not found on 16 Nov 2026, so the report is the first source here. ## Sources - [Harmonic Security Raises $50 Million for Workforce AI Protection, SecurityWeek](https://www.securityweek.com/harmonic-security-raises-50-million-for-workforce-ai-protection/) (securityweek.com), read 2026-11-16 - [Harmonic Security raises $50M Series B to expand US and UK teams, Unite.AI](https://www.unite.ai/harmonic-security-raises-50m-series-b-to-expand-us-and-uk-teams/) (unite.ai), read 2026-11-16 ## Related pages - [Harmonic Security profile](/vendors/harmonic-security/) - [All news](/news/) - [Zscaler adds MCP server policies to AI Asset Management](/news/zscaler-ai-asset-management-adds-mcp-server-policies/) - [Aurascape adds masking and model-level policy](/news/aurascape-september-release-masking-and-model-level-policy/) - [The ranking](/ranking/) --- # Aurascape adds masking and model-level policy URL: https://shadowaidetectiontools.com/news/aurascape-september-release-masking-and-model-level-policy/ Updated: 2026-10-07 Aurascape described its September 2026 release in a post dated 7 October 2026. Masking, a mode of the redact action, replaces a sensitive value in a prompt before it reaches the model and restores it in the reply. Policy can also allow a platform such as Amazon Bedrock while blocking a named model inside it. 2026-10-07 Aurascape's product update, written by its product marketing manager and dated 7 October 2026, introduces sensitive data masking as a mode of the redact action in its Protection policy. The model works with a placeholder and Aurascape restores the real value in the response the user reads. Administrators can unmask a value in the Conversations view, depending on role permissions. The post also deepens model-level policy. Its example is allowing Amazon Bedrock while blocking DeepSeek within it. Aurascape says model-level policies can warn, ask for confirmation or block use of a restricted model across services. On-demand reports gain Unique User Counts by Application and a Top Protection Policy Hits chart, and administrators get more control over notification delivery and data classification settings. The post gives no price and no customer figures. ## Sources - [September New Innovations Recap: More Precise Protection for AI Work, Aurascape](https://aurascape.ai/resources/blog/september-2026-product-update/) (aurascape.ai), read 2026-11-16 ## Related pages - [Aurascape profile](/vendors/aurascape/) - [All news](/news/) - [Harmonic Security raises $50 million Series B](/news/harmonic-security-raises-50-million-series-b/) - [Zscaler describes Endpoint AI Security for shadow AI](/news/zscaler-endpoint-ai-security-discovers-ai-on-managed-devices/) - [The ranking](/ranking/) --- # Zscaler describes Endpoint AI Security for shadow AI URL: https://shadowaidetectiontools.com/news/zscaler-endpoint-ai-security-discovers-ai-on-managed-devices/ Updated: 2026-10-05 Zscaler's 5 October 2026 blog post describes Endpoint AI Security for managed devices. Zscaler says it identifies AI assistants, agents, browsers, development tools, models, packages, extensions and connected services, and can restrict unauthorized agents and tools. A Claude Mythos 5.1 feature, Exploit Paths, is in limited preview. 2026-10-05 Zscaler published a product blog on 5 October 2026 about Endpoint AI Security. It frames shadow AI as a growing layer of applications, agents, extensions, models, packages and configuration files on employee devices that falls outside normal software rollout. Zscaler says the product combines three things: discovery of AI assets on managed devices, runtime visibility into prompts, tool calls, file access, commands and network activity, and a policy layer that can restrict unauthorized agents and tools, stop dangerous commands and protect sensitive files. The post also announces limited preview availability of Anthropic's Claude Mythos 5.1 inside the product, powering a feature called Exploit Paths that connects related endpoint AI activity. The blog states no price and no customer count. ## Sources - [Bringing Anthropic's Claude Mythos 5.1 to Zscaler Endpoint AI Security](https://www.zscaler.com/blogs/product-insights/bringing-anthropic-s-claude-mythos-5-1-zscaler-endpoint-ai-security) (zscaler.com), read 2026-11-16 ## Related pages - [Zscaler profile](/vendors/zscaler/) - [All news](/news/) - [Aurascape adds masking and model-level policy](/news/aurascape-september-release-masking-and-model-level-policy/) - [Reco raises $55 million as it widens from SaaS to agents](/news/reco-raises-55-million-agent-security/) - [The ranking](/ranking/) --- # Reco raises $55 million as it widens from SaaS to agents URL: https://shadowaidetectiontools.com/news/reco-raises-55-million-agent-security/ Updated: 2026-09-29 Reco announced a $55 million raise on 29 September 2026, TechCrunch reported, bringing total capital raised to $140 million. TechCrunch also reports more than 100 customers, integrations with more than 280 apps, and AT&T as a customer investor. Reco says its platform found 21,000 agents that one Fortune 100 customer did not know about. 2026-09-29 TechCrunch reported on 29 September 2026 that Reco raised $55 million, building on a $30 million Series B announced in February 2026. AT&T, a customer, invested through its venture arm, as did Forestay and Quadrille Capital. The new funding brings Reco's total capital raised to $140 million, per TechCrunch. The article says Reco moved from mapping and securing SaaS and AI platforms to a wider agent security product built on a context graph. Reco's chief executive told TechCrunch that at one Fortune 100 customer the platform found 21,000 agents the company did not know about. That is a vendor claim reported by the outlet; no method or sample is given. TechCrunch reports Reco integrates with more than 280 apps and uses browser and network signals to find agents outside apps it connects to directly. Annual recurring revenue is described by the CEO as double-digit millions of dollars. This item rests on TechCrunch's reporting; a Reco release page for the round was not found on 16 Nov 2026. ## Sources - [Reco raises $55M as AI agent security startups crowd the market, TechCrunch](https://techcrunch.com/2026/09/29/reco-raises-55m-as-ai-agent-security-startups-crowd-the-market/) (techcrunch.com), read 2026-11-16 ## Related pages - [Reco profile](/vendors/reco/) - [All news](/news/) - [Zscaler describes Endpoint AI Security for shadow AI](/news/zscaler-endpoint-ai-security-discovers-ai-on-managed-devices/) - [Zscaler adds AI Security policies for cloud agent platforms](/news/zscaler-adds-ai-security-policies-for-bedrock-foundry-copilot-studio/) - [The ranking](/ranking/) --- # Zscaler adds AI Security policies for cloud agent platforms URL: https://shadowaidetectiontools.com/news/zscaler-adds-ai-security-policies-for-bedrock-foundry-copilot-studio/ Updated: 2026-08-19 Zscaler's AI Asset Management release notes record new AI Security policies deployed on 19 August 2026. The policies cover AWS Bedrock agents, Azure AI Foundry Hub agents, Azure OpenAI agents and Microsoft Copilot Studio agents, and they flag settings such as disabled prompt attack filters and agents using publisher dev credentials. Read 16 Nov 2026. 2026-08-19 Zscaler's release notes for AI Asset Management say that on 19 August 2026 a set of new AI Security policies became available. Titles listed include an AWS Bedrock agent with a prompt attack filter disabled, an AWS Bedrock agent with confirmation for function-based action groups disabled, and an AWS Bedrock agent with code interpretation enabled without guardrails. The same entry lists policies for Azure AI Foundry Hub agents (Bing search enabled, empty instructions, tools that use API keys in security schemes), for Azure OpenAI agents (empty instructions) and for Microsoft Copilot Studio agents (manual authentication configured, use of the publisher's dev credentials). For a shadow AI buyer the change is about the catalog of what gets checked once an AI asset is found: the policies read the configuration of agents already built in the company's cloud accounts, not the web apps staff open in a browser. The release notes give no customer counts and no pricing. ## Sources - [Release Upgrade Summary (2026), Zscaler AI Asset Management](https://help.zscaler.com/ai-asset-mgmt/release-upgrade-summary-2026) (help.zscaler.com), read 2026-11-16 ## Related pages - [Zscaler profile](/vendors/zscaler/) - [All news](/news/) - [Reco raises $55 million as it widens from SaaS to agents](/news/reco-raises-55-million-agent-security/) - [Nudge Security adds OpenAI Agents connected app](/news/nudge-security-adds-openai-agents-connected-app/) - [The ranking](/ranking/) --- # Shadow AI detection tools: frequently asked questions URL: https://shadowaidetectiontools.com/faq/ Updated: 2026-11-16 Answers to common questions about shadow AI detection tools, including whether a CASB is enough, whether detection works without an agent, what Microsoft Purview sees, whether AI note takers show up, what tools cost and how often to refresh the inventory. Facts were read on 16 Nov 2026. ## Questions and answers ### What is a shadow AI detection tool? A shadow AI detection tool finds the AI apps, agents, browser extensions, IDE plugins and MCP servers that staff use without approval. It reads one or more sources, such as the endpoint, browser, network traffic, the identity provider or SaaS APIs, and builds an inventory. Many also warn, redact, block or revoke access when they find something. Source: [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16. ### Do I need a shadow AI detection tool or is a CASB enough? A CASB or cloud discovery feature finds web AI apps that appear in network traffic and match its catalog, which is a useful start. It misses local coding agents, extensions, MCP servers and AI features inside approved apps. Microsoft's tutorial states that apps outside the catalog are not discovered by default, so add an endpoint, browser or SaaS API source. Source: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. ### Can shadow AI be detected without installing an agent? Yes, from some vantage points. Nudge Security says it needs no endpoint agents or network proxies and starts from one read-only email API connection. Grip Security describes agentless API connections to email, identity and SaaS, and Palo Alto Networks AI Access Security runs through existing gateways. Agentless sources miss local apps, coding agents and MCP servers on devices. Sources: [Nudge Security: AI security use case](https://www.nudgesecurity.com/use-cases/ai-security-3), read 2026-11-16; [Grip Security source (grip-security-10)](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16; [Grip Security source (grip-security-12)](https://www.grip.security/okta-integration), read 2026-11-16; [Palo Alto Networks source (palo-alto-networks-14)](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16. ### Does Microsoft Purview detect shadow AI? Partly. Microsoft Purview says AI apps seen through browser activity are filed as Generative AI in the Defender for Cloud Apps catalog, and DSPM for AI needs the Purview browser extension and onboarded devices. Policies need at least 24 hours to collect data. Local agent discovery sits in Defender for Endpoint, documented for Windows and macOS. Source: [Data Security Posture Management for AI, Microsoft Purview](https://learn.microsoft.com/en-us/purview/dspm-for-ai), read 2026-11-16. ### Do detection tools see AI note takers? It depends on the source. A note taker that is granted access to a calendar or mailbox shows up as an OAuth grant, which identity and SaaS tools such as Grip Security, Reco, Nudge Security and Push Security list. A bot that joins meetings through a web link may show only in network or browser records. No tool sees every case. Sources: [Grip Security: discover SaaS risks](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16; [Reco source (reco-03)](https://www.reco.ai/platform/application-discovery), read 2026-11-16; [Nudge Security source (nudge-security-05)](https://www.nudgesecurity.com/post/how-to-monitor-remote-mcp-server-connections-with-nudge-security), read 2026-11-16; [Push Security source (push-security-05)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. ### How much do shadow AI detection tools cost? Few vendors publish prices. Push Security lists $5 per user per month for up to 500 employees, and Nudge Security lists $750 per month for up to 150 users, then $5 per user per month from 150 to 1,500. Microsoft lists Agent 365 at $15.00 per user per month. Most others publish none: none found on 16 Nov 2026. Sources: [Nudge Security pricing](https://www.nudgesecurity.com/pricing), read 2026-11-16; [Push Security pricing](https://pushsecurity.com/pricing), read 2026-11-16; [Microsoft Agent 365](https://www.microsoft.com/en-us/microsoft-agent-365), read 2026-11-16. ### Is there an open-source shadow AI detector? No product among the 16 profiled here is documented as an open-source detector, as read on 16 Nov 2026. A free starting point is the data you already hold: Entra ID sign-in and audit logs, Google Workspace API controls, and firewall or proxy logs. Microsoft documents a manual snapshot report that accepts proxy logs for a first view of app usage. Source: [Create snapshot cloud discovery reports, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/create-snapshot-cloud-discovery-reports), read 2026-11-16. ### How often should the AI inventory be refreshed? Refresh it continuously from automated sources and review the whole inventory at least quarterly. New AI tools appear weekly, and a quarterly snapshot alone misses tools that spread between reviews. Discovery policies that alert when a new app is detected, such as the Microsoft Defender for Cloud Apps app discovery policy, keep the inventory current between reviews. Source: [Create cloud discovery policies, Microsoft Defender for Cloud Apps](https://learn.microsoft.com/en-us/defender-cloud-apps/cloud-discovery-policies), read 2026-11-16. ### Does Zscaler detect shadow AI? Yes. Zscaler documents AI Asset Management with separate public AI inventories for GenAI web apps, embedded AI and endpoint AI, plus an MCP server inventory. It reads inline traffic through the Zero Trust Exchange and adds an endpoint layer. None found on 16 Nov 2026 for a published price or a time to a first inventory. Source: [Zscaler: public AI inventory, GenAI](https://help.zscaler.com/ai-asset-mgmt/about-public-ai-inventory-gen-ai), read 2026-11-16. ### Can a browser extension detect shadow AI? Yes, for web use. A browser extension sees which AI sites a person signs in to, which account type is used and which AI extensions are installed. Push Security, Nudge Security, Grip Extend and LayerX work this way. An extension does not see desktop AI apps, command-line agents or local MCP servers, which need an endpoint source. Sources: [Push Security: shadow AI, how to discover, govern and secure AI apps](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16; [Nudge Security source (nudge-security-15)](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16; [Grip Security source (grip-security-10)](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16; [LayerX (Akamai Workforce Protector) source (layerx-12)](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16. ### Can shadow AI tools block unapproved AI instead of only reporting it? Many can. Netskope, Zscaler, Aurascape, Cyberhaven, Palo Alto Networks, Harmonic and Push Security document blocking, alongside warn, coach and redact actions. Nudge Security nudges rather than blocks, and Reco is described by The Hacker News as detection and alerting that cannot prevent connections. The Finding-to-Action Ladder on this site shows each action. Sources: [Palo Alto Networks: discover risks posed by GenAI apps](https://docs.paloaltonetworks.com/ai-access-security/administration/discover-risks-posed-by-genai-apps), read 2026-11-16; [Netskope source (netskope-18)](https://www.netskope.com/products/ai-products), read 2026-11-16; [Zscaler source (zscaler-05)](https://www.zscaler.com/products-and-solutions/ai-security), read 2026-11-16; [Aurascape source (aurascape-21)](https://aurascape.ai/frictionless-ai-security/), read 2026-11-16; [Cyberhaven source (cyberhaven-17)](https://www.cyberhaven.com/product/ai-security), read 2026-11-16; [Palo Alto Networks source (palo-alto-networks-20)](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16; [Harmonic Security source (harmonic-security-28)](https://www.harmonic.security/pricing), read 2026-11-16; [Push Security source (push-security-17)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16; [Nudge Security source (nudge-security-19)](https://www.nudgesecurity.com/features/browser-extension), read 2026-11-16; [Reco source (reco-10)](https://thehackernews.com/2025/01/product-review-how-reco-discovers.html), read 2026-11-16. ### Which tools find MCP servers? Microsoft Defender for Endpoint reports each discovered local agent's configured MCP servers. Zscaler keeps an MCP server inventory, and Harmonic, Netskope, Cyberhaven, Aurascape, Aim Security, WitnessAI, SentinelOne and Bay list MCP servers in their documentation or claims. Grip Security, Reco, LayerX and Palo Alto Networks show none found on 16 Nov 2026 for MCP server discovery. Sources: [Discover local AI agents, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/discover-local-ai-agents), read 2026-11-16; [Zscaler source (zscaler-13)](https://help.zscaler.com/ai-asset-mgmt/about-ai-assets-mcp-servers), read 2026-11-16; [Harmonic Security source (harmonic-security-12)](https://www.harmonic.security/solutions/endpoint-ai-security), read 2026-11-16; [Netskope source (netskope-06)](https://www.netskope.com/products/ai-command-center), read 2026-11-16; [Cyberhaven source (cyberhaven-03)](https://www.cyberhaven.com/press-releases/cyberhaven-solves-the-surge-in-shadow-ai), read 2026-11-16; [Aurascape source (aurascape-08)](https://aurascape.ai/secure-agentic-ai/), read 2026-11-16; [Aim Security source (aim-security-13)](https://www.catonetworks.com/solutions/ai-security/), read 2026-11-16; [WitnessAI source (witnessai-08)](https://witness.ai/for-employees/), read 2026-11-16; [SentinelOne (Prompt Security) source (sentinelone-10)](https://www.sentinelone.com/platform/prompt-security-agentic-ai-security/), read 2026-11-16; [Bay source (bay-11)](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=faq), read 2026-11-16; [Grip Security source (grip-security-09)](https://www.grip.security/mcp-security), read 2026-11-16; [Reco source (reco-09)](https://www.reco.ai/use-cases/shadow-ai-discovery), read 2026-11-16; [LayerX (Akamai Workforce Protector) source (layerx-11)](https://www.akamai.com/products/akamai-workforce-protector), read 2026-11-16; [Palo Alto Networks source (palo-alto-networks-11)](https://docs.paloaltonetworks.com/ai-access-security), read 2026-11-16. ### How is shadow AI different from shadow IT? Shadow IT is an unapproved app or service, usually seen as a new domain. Shadow AI adds local agents, extensions, MCP servers and AI features switched on inside approved apps, which leave no new domain. Detection therefore needs endpoint, identity and SaaS sources in addition to network traffic and an app catalog. Source: [Microsoft Defender for Cloud Apps: discover and manage shadow IT](https://learn.microsoft.com/en-us/defender-cloud-apps/tutorial-shadow-it), read 2026-11-16. ### Does Microsoft Defender find coding agents on laptops? Yes, on supported devices. Microsoft says Defender for Endpoint discovers local AI agents, including coding agents, IDE extensions and desktop assistants, with no extra deployment beyond onboarding. Its support matrix names Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf and Kiro, and discovery is documented for Windows and macOS. Source: [AI agent support matrix, Microsoft Defender for Endpoint](https://learn.microsoft.com/en-us/defender-endpoint/ai-agent-support-matrix), read 2026-11-16. ### Can shadow AI be revoked after an app is granted OAuth access? Yes, in the identity provider and in some detection tools. Entra ID lets admins review and revoke permissions granted to enterprise applications, and Google Workspace API controls can set an app to Blocked. Grip Security says it can revoke access, Nudge Security lists auto-revoking permissive grants, and Palo Alto Networks documents revoking by privilege scope. Push Security discovers AI OAuth grants; revocation: none found on 16 Nov 2026. Sources: [Review permissions granted to enterprise applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions), read 2026-11-16; [Grip Security source (grip-security-07)](https://www.grip.security/platform/discover-saas-risks), read 2026-11-16; [Nudge Security source (nudge-security-16)](https://www.nudgesecurity.com/pricing), read 2026-11-16; [Palo Alto Networks source (palo-alto-networks-18)](https://www.paloaltonetworks.com/sase/ai-access-security), read 2026-11-16; [Push Security source (push-security-21)](https://pushsecurity.com/blog/shadow-ai-how-to-discover-govern-and-secure-ai-apps), read 2026-11-16. ### What is the difference between shadow AI and agentic AI? Shadow AI describes approval status: AI in use without sanction. Agentic AI describes autonomy: software that takes actions on its own. A coding agent a developer installs without approval is both. That is why discovery tools list agents and MCP servers next to chat apps, and some, such as Bay, give each tool an autonomy score. Source: [Bay: shadow AI discovery](https://www.bay.io/shadow-ai-discovery?utm_source=shadowaidetectiontools.com&utm_medium=referral&utm_campaign=faq), read 2026-11-16. ## Related pages - [Guides](/guides/) - [Glossary](/glossary/) - [The ranking](/ranking/) - [Method](/method/) --- # About Shadow AI Detection Tools URL: https://shadowaidetectiontools.com/about/ Updated: 2026-11-16 Shadow AI Detection Tools publishes a ranked comparison of 16 products that find the AI apps, agents, plugins and extensions staff use without approval. Every score comes from a sourced fact file read on 16 Nov 2026, weighted by criteria committed before any score existed. ## What does this publication cover? It covers tools that discover unapproved AI: web apps, desktop apps, coding agents, browser extensions, IDE plugins, MCP servers and AI features embedded in approved SaaS. The reader is a security lead who must find the AI already in use. The site publishes the ranking, five situation views, 16 vendor profiles, two downloadable datasets (the Vantage Point Matrix and the Finding-to-Action Ladder), guides, a glossary, dated news and an FAQ. ## How is the work organized? Facts are gathered once per vendor, quote-checked, and scored blind against fixed weights. The method page describes each step, and every page cites the sources behind its claims with the date read. The site is a static publication. Numbers on ranked pages are generated from the published scores files, so a page and its data cannot disagree. ## Related pages - [Method](/method/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [The ranking](/ranking/) - [FAQ](/faq/) --- # Editorial policy URL: https://shadowaidetectiontools.com/editorial-policy/ Updated: 2026-11-16 Every page follows one set of rules: each fact carries a source URL and read date, vendor claims are written as the vendor's, scores follow weights fixed on 16 Nov 2026, and a score changes only when a sourced fact changes. Errors are corrected in public with a dated log entry. ## What are the sourcing rules? Every fact carries a source URL and the date it was read. A vendor's claim is written as the vendor's, such as Harmonic says or per Microsoft's docs, and a claim scores at most 7 until a second source confirms it. Where nothing is found the page says none found on the read date. Security content covers risks, detection, policy and fixes only, each sourced, with no exploit code and no steps to bypass a control. ## How are scores kept consistent? The same research plan runs for every vendor, the same fixed weights apply, and scoring is blind to vendor identity. Each vendor's strength is stated as the evidence shows it, and no competitor fact is dropped to change an order. A score changes only when a sourced fact changes, and the change is logged with the fact and date. ## How are errors handled? Corrections are accepted from anyone with a source, through the process on the corrections page. A confirmed error is fixed on the page, in the data files and in the changelog on the same day where possible. ## Related pages - [Method](/method/) - [Corrections](/corrections/) - [The ranking](/ranking/) - [FAQ](/faq/) --- # Corrections and tips URL: https://shadowaidetectiontools.com/corrections/ Updated: 2026-11-16 To report an error on Shadow AI Detection Tools, email editor@shadowaidetectiontools.com with the page, the sentence or number, and a source URL that shows the fact. Confirmed corrections are applied with a dated entry. The same address takes tips about product changes, releases and documentation updates. ## How do I report an error? Email [editor@shadowaidetectiontools.com](mailto:editor@shadowaidetectiontools.com?subject=Correction) with the page address, the exact sentence or figure, what is wrong, and a source URL that shows the correct fact. A vendor primary source, such as a documentation page or release note, is the quickest to verify. Each report is checked against the source. If the fact changes, the page, the data and the changelog are updated with the date, and a score changes only when the fact it rests on changes. ## Send a tip Tips are welcome at [editor@shadowaidetectiontools.com](mailto:editor@shadowaidetectiontools.com?subject=Tip): a new product release, a changed price, a vendor documentation page that went public, or a study on unapproved AI use with the sample size stated. Include the URL. Tips are treated as leads, not facts. A tip is read, fetched and quote-checked before anything is published. ## What is logged? Corrections that change a fact or a score are recorded in the changelog on the method page with the date, the old and new values and the source. Typo fixes are not logged. ## Write to the editor Email [editor@shadowaidetectiontools.com](mailto:editor@shadowaidetectiontools.com) with the page address, what is wrong and the source that shows it. ## Related pages - [Method](/method/) - [Editorial policy](/editorial-policy/) - [The ranking](/ranking/) - [FAQ](/faq/) --- # Contact URL: https://shadowaidetectiontools.com/contact/ Updated: 2026-11-16 Email editor@shadowaidetectiontools.com for corrections, tips, vendor evidence or questions about the method. Include the page address and, for a factual change, a source URL. Messages about errors are answered with the outcome of the check and the date any change was made. ## What can I write about? The address is [editor@shadowaidetectiontools.com](mailto:editor@shadowaidetectiontools.com). Useful messages name a page, quote the sentence in question and link a primary source. There is no contact form and no tracking on this site; email is the only channel. ## Write to the editor Email [editor@shadowaidetectiontools.com](mailto:editor@shadowaidetectiontools.com). ## Related pages - [Method](/method/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [The ranking](/ranking/) - [FAQ](/faq/) --- # Privacy policy URL: https://shadowaidetectiontools.com/privacy/ Updated: 2026-11-16 Shadow AI Detection Tools is a static site that sets no cookies, runs no analytics scripts and loads no third-party trackers. Amazon CloudFront access logs record request details such as IP address, time, page and user agent, and those logs are kept for 90 days. Email sent to the editor is used to answer it. ## What data is collected? The server records standard access logs through CloudFront: IP address, time, requested path, status, referrer and user agent. They are kept for 90 days to measure load and diagnose faults, then deleted. The site sets no cookies, has no analytics scripts, no advertising tags, no forms and no accounts. ## What happens to email? A message to the editor is read to answer it and kept only as long as the correction or tip needs a record. A corrections log entry names the page and the source, not the sender, unless the sender asks to be credited. ## Can I ask about my data? Send a request to the editor address, naming the date and approximate time of the visit. IP addresses in logs are matched to the request and deleted on request where the log still exists. ## Related pages - [Method](/method/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [The ranking](/ranking/) - [FAQ](/faq/) --- # Terms of use URL: https://shadowaidetectiontools.com/terms/ Updated: 2026-11-16 Shadow AI Detection Tools content and datasets are licensed under CC BY 4.0: reuse is allowed with attribution using the citation line on the datasets page. Rankings reflect the evidence read on 16 Nov 2026 and the weights published on the method page. Each fact carries its source and read date so a reader can check it. ## How may the content be reused? Text, scores and datasets may be copied, shared and adapted under CC BY 4.0 with attribution. Use the citation line on the datasets page: title, edition date, URL and license. ## How current are the facts? Every fact is as of the read date printed beside its source, 16 Nov 2026 for this edition. Detection products add catalogs, connectors and actions often, so the linked source page is the place to check a fact again. Links to external sources are there so readers can check each fact against the page it came from. ## How is an error reported? Use the corrections page. A report that cites a public source is checked against it, and a confirmed error is fixed and logged with its date. ## Related pages - [Method](/method/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [The ranking](/ranking/) - [FAQ](/faq/)