Token Security report: 82% of firms find unknown AI agents despite confidence in visibility
A Cloud Security Alliance survey of 418 IT and security professionals, commissioned by Token Security in January 2026, found that while 68% believed they had strong visibility into their organization's AI agents, 82% had discovered unsanctioned agents within the past year. Token Security illustrates the risk with a hypothetical case: an employee building a simple AI agent using OAuth consent and a personal API key, entirely outside security team oversight.
GoKawiil's interpretation of the reporting above, not reported fact.
The gap between perceived and actual visibility suggests many enterprises may be underestimating how easily employees can spin up AI agents with access to sensitive systems using no-code tools and personal credentials. This could mean security teams are reacting to shadow AI deployments after the fact rather than preventing them, especially as access methods like OAuth grants and API keys make agent creation nearly frictionless.
- 68% of surveyed professionals felt confident in their AI agent visibility, yet 82% found unknown agents in the past year.
- The survey covered 418 IT and security professionals and was commissioned by Token Security via the Cloud Security Alliance.
- Agents can gain access to sensitive systems like ERPs and Slack through personal API keys and OAuth consent screens without security team involvement.
Source: bleepingcomputer.com, 2026-11-13
Published there as: “How one AI agent exposed five critical enterprise security gaps”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.