Thursday, October 29, 2026 · Claude & Anthropic
Claude · Safety · happened Oct 27, 2026
Anthropic's red team says an unreleased Claude model chained flaws into working exploits for Chrome, Safari, Android, Windows and Linux
Anthropic's Frontier Red Team published five exploit chains that an unnamed internal research model built in the lab, including a Chrome sandbox escape and remote root on some Linux servers. Anthropic says the flaws were reported to vendors and fixed first; The Hacker News notes the post doesn't say how much of each chain the model did without human help, and that only some fixes publicly credit Anthropic.
Why it matters
It is another sign that frontier models can now turn bugs into real attacks, which raises the stakes for how such models are released.