By Vero SecurityBuilt and funded by Vero Security, the AI Security Engineering platform.Read the charter

Excessive agency

Excessive agency means an agent holds more tools, permissions, or autonomy than its task needs, so any manipulation or mistake has a large blast radius. Many public agent incidents pair a prompt injection with this kind of overreach rather than relying on a clever exploit.

Read in this order

  1. OfficialLLM06:2025 Excessive AgencyStart here: the standard definition, causes, and mitigations.
  2. VideoDEF CON 34 - Main - Escaping Copilot's AI Guarded Sandbox - Simon Maxwell-Stewart, Ryan HausknechtReconstructs a Copilot Studio code-interpreter escape across three controls. 38 min.
  3. ArticleAnatomy of a frontier lab agent intrusion: A technical timeline of the July 2026 incident (Hugging Face)Evaluation agents with too much reach left their range and hit a real production system.

Save your progress

We email you a sign-in link. No password.

By signing in you agree to the terms. The privacy policy explains how we use your details.

Join the AI security engineering community

Free. Roadmap updates, chapters near you, and first word when credentials open.

We’ll email you community updates. Unsubscribe anytime. By sending this you agree to the terms. The privacy policy explains how we use your details.