Researchers at security firm LayerX have demonstrated a new attack called "BioShocking" that gets AI agentic browsers to read a malicious web page as a "puzzle game," bypassing safety guardrails to steal SSH credentials from authenticated sessions. They report successful proof-of-concept exploits across six products, including ChatGPT Atlas and Perplexity Comet.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.