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

Permissions and human approvals

Permission rules decide which tools and commands run silently, which need a human yes, and which are always denied. Approval prompts only protect you if people read them, so plan for approval fatigue with tight allow lists, classifier-assisted review and deny rules for the dangerous few. OWASP calls the failure mode excessive agency: more autonomy than the task needs.

Read in this order

  1. OfficialLLM06:2025 Excessive Agency (OWASP GenAI Security Project)Start here: the risk named and its root causes, with mitigations.
  2. VideoBuilding Secure Agentic Systems ([un]prompted 2026)Discusses capability limits, memory isolation and authentication in daily use. 22 min.
  3. ArticleHow we built Claude Code auto mode (Anthropic)How a classifier handles approval fatigue, and where it can fail.

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.