News Summary

Anthropic has launched a new AI Chrome extension, Claude for Chrome, which introduces a security challenge as AI assistants become capable of controlling web browsers. The extension allows users to chat with the Claude AI model in a sidebar and can perform tasks like managing calendars and drafting emails. However, it poses a security risk as AI browser agents can be tricked into harmful actions nearly a quarter of the time through prompt-injection attacks. This vulnerability highlights the competitive rush among AI labs to integrate AI into browsers, with companies like Perplexity and OpenAI launching similar AI agents. Despite extensive testing by Anthropic, the security flaws remain a significant concern.