Shadow AI
AI tools and agents that employees adopt without the security team knowing. It is the AI-era version of shadow IT.
In our briefings
The claim each of the newest briefings on Shadow AI would have you leave with.
Mattson, SecureAuth's chief executive, argues that a model cannot be secured the way tested software is: it is non-deterministic, and a request phrased another way, or in another language, gets around whatever it was told not to do. His advice is to "just assume that it could do anything any time" and put the guardrails around the model, so that a financial transaction or a database change the agent attempts is approved and recorded. He cites Anthropic as saying no guardrail is adequate. The statement we take him to mean, Anthropic's of 12 June, which he recalls as "three weeks ago, I believe", says the company suspects perfect jailbreak resistance "is not currently possible for any model provider", while defending its own safeguards as strong. Briefings we have published on other sources move the control off the model too: a Silverfort guest calls guardrails on a probabilistic system guidance rather than control, ClearVector's John Laliberte compares them to bowling bumpers, and a zero-trust framework paper for agents adds a session-enforcement layer.
When the model cannot be trusted, control what the agent can reach, checked on every actionBI-FIND-2026-0151
On a sponsored Packet Protector episode, host Jennifer Jabush cited Palo Alto Networks' latest Identity Security Landscape report: machine identities "outnumber humans 109 to 1". CyberArk's Uzi Alon, whose company is now part of Palo Alto, said the yearly study put the ratio at "about 50" two years ago and 82 last year. 82 to 1 is also the ratio in CyberArk's 2025 release, by our reading of it. The ratio comes from the vendor that sells the fix. Neither speaker says how it was counted. Our briefing on the report found that its public excerpt thanks "2,930 cybersecurity decision-makers worldwide" but prints no question, base or counting unit for the ratio. The host also cited the report for over 40 percent of AI agents having direct access to critical enterprise data. The guest did not repeat that number.
Palo Alto's own survey puts machine identities at 109 per person, up from 82 a year earlierBI-FIND-2026-0719
In Microsoft's scripted Ignite demo, the AI gateway in Entra Suite blocked three things in network traffic: an upload of customer data to ChatGPT, under a Purview file policy; a prompt injection hidden in a document as part of the demo's story; and an AI assistant's connection to an unsanctioned MCP server. Each block was logged. Igor Sakhnov called the gateway "industry first" and said it "works at the network level" with no integration into AI apps. He also said he could build an MCP server in ten minutes and asked how much thought for authorization would fit in that time. We read the control as deciding which MCP servers an assistant can reach, not what an approved one may do, and Microsoft gave no rate for missed injections or wrongly blocked traffic. The presenters called the features they showed just announced or recently released, but gave no preview or general-availability status and no date for any of them.
Microsoft's Entra Suite demo blocks a planted prompt injection and an unsanctioned MCP server but gives no catch rateBI-FIND-2025-0541
›Cited in 11 briefings
Every briefing that discusses this term, newest first.
- When the model cannot be trusted, control what the agent can reach, checked on every action
Podcast · · Tech Talks Daily · AI Agent Security: Why Identity and Access Control Matter More Than Guardrails · Aug 11, 2026
- Palo Alto's own survey puts machine identities at 109 per person, up from 82 a year earlier
Podcast · · Packet Protector · PP115: Palo Alto Networks: Reality of 109 to 1: Securing Machine Identities and AI Agents (Sponsored) · Jun 23, 2026
- Microsoft's Entra Suite demo blocks a planted prompt injection and an unsanctioned MCP server but gives no catch rate
Video · · Microsoft Events · Uploaded November 24, 2025
- Textile maker's CIO and CISO has no way to find shadow AI, and calls an agent its own kind of identity
Podcast · · Identity at the Center · #446 - Rethinking Identity for AI Agents with Rick Scot · Sep 7, 2026
- Mythos also reports bugs that are not there, says Trend AI's CISO, so the work is in checking them
Podcast · · Cloud Security Podcast · Mythos Hype vs. Reality: Why You Need an Agentic Harness · Oct 21, 2026
- Okta says its agent kill switch revokes all tokens; its product screen says tokens outside the gateway survive
Video · · Okta · Sep 24, 2026
- Okta's Oktane keynote: two agent products generally available now, support for other identity providers in 2027
Video · · Okta · Sep 23, 2026
- The directors' own handbook now puts one question about AI agent identity to boards
Article · · National Association of Corporate Directors and Internet Security Alliance · Larry Clinton et al. · Mar 12, 2026
- Ping moved agent identity from roadmap to shipping product, and let AI agents run its admin tools too
Conference · · Ping Identity · Austin · 2-day event (Day 1 / Day 2), 2026
- The board's own handbook now asks about AI agent identity by name, and the first answer is an inventory
Podcast · · The Non-Human & AI Identity Podcast · NHI & AI Identity Summit : The CISO Imperative · Aug 28, 2026
- Once people hand their credentials to an AI, the question is whether the account is still human
Podcast · · Nexus: A Claroty Podcast · John Laliberte on AI and Non-Human Identity Controls · Aug 5, 2026
Related terms
What appears alongside Shadow AI in the same briefings.