New Bio Shocking Attack Tricks AI Browsers Into Leaking User Credentials
Summary: Security firm LayerX developed a technique called BioShocking that tricked six AI browsers and assistants - including OpenAI's ChatGPT Atlas, Perplexity's Comet, and Anthropic's Claude browser extension - into copying a user's credentials and sending them to an attacker. The attack uses a puzzle web page that rewards wrong answers, getting the agent to follow game logic instead of safety logic until it pulls SSH credentials from a work GitHub repository and reports the theft as a win.
Key Takeaway: Not one of the six agents refused the final step, and when LayerX reported the issue between October 2025 and January 2026, Perplexity closed the report without acting, Fellou, Genspark, and Sigma did not respond, and Anthropic's Claude fix did not hold - leaving an agent in agent mode as another account with reach into company systems.
Read the original at The Hacker News ↗
- Topics
- AI Security
- Related
- SECURITY AFFAIRS AI-CYBERSECURITY NEWSLETTER ROUND 8 Cyber Risks · Nov 15, 2026Pentagon Renames CDAO ‘Super Intelligence Office’ Cyber Risks · Nov 15, 2026186 Fake AI Domains Redirect Users to Click Fix Lures, Browser Malware, and Search Hijackers Cyber Risks · Nov 12, 2026