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

Monitoring agents for misbehavior

Research on AI control and sabotage evaluation asks whether a trusted monitor can catch an agent doing something it should not. Vendor threat reports show what real misuse looks like in telemetry, which is useful when tuning alerts.

Read in this order

  1. ArticleSHADE-Arena: Evaluating sabotage in LLM agents (Anthropic)Start here: test results on whether monitors catch agent sabotage.
  2. PaperAI Control: Improving Safety Despite Intentional SubversionThe control protocols that frame monitoring as a security problem.
  3. ArticleDetecting and countering misuse of AI: August 2025 (Anthropic)Real misuse cases and how they were detected.

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.