Article

Claude in Chrome is generally available

Anthropic · claude.com

Anthropic on letting Claude act in the browser without approving each step, with a safety classifier checking every action.

Open on claude.com ↗

It answers 2 questions

  1. What can go wrong when an agent acts on my behalf, and how do I keep it safe?
  2. What can AI agents and AI browsers (ChatGPT Work, Comet, Claude in Chrome) do for my research, and are they worth it?

Read with it

  1. The single most important security idea for anyone deploying agents. simonwillison.net Article The lethal trifecta for AI agents Simon Willison
  2. The same argument in shareable form for your team channel. simonw.substack.com Article The lethal trifecta (newsletter edition) Simon Willison
  3. The full running history of the attack class, from the person who named it. simonwillison.net Article Prompt injection series Simon Willison
  4. The foundational explainer that makes the attack click in 15 minutes. simonw.substack.com Video Prompt injection explained, with video, slides, and a transcript Simon Willison
  5. Fresh real-world exploits showing this is not theoretical. simonw.substack.com Article Prompt injections as far as the eye can see Simon Willison
  6. The industry-standard checklist, free and framework-agnostic. cheatsheetseries.owasp.org Guide AI Agent Security Cheat Sheet OWASP
eChai Partner Brands The partnership →