Claude AI Builds One-Click Chrome and Zero-Click Linux Exploit Chains, Bypasses Android Lock Screen

What happened?

Anthropic’s Frontier Red Team says an unreleased Claude research model wrote working exploit chains for Chrome, Safari, Android, Windows, and Linux. The company shared the details in on research post on October 27, 2026. These were lab tests and not real-world incidents. Anthropic reported every bug to the vendor first, and the vendor released fixes. Chrome Falls to a One-Click Chain The Chrome exploit is the most detailed. After a first prompt, Claude wrote a chain that runs code outside the browser sandbox.

Key details

• Anthropic's Claude research model generated exploit chains for Chrome, Safari, Android, Windows, and Linux in lab tests.
• Chrome exploit used four bugs: V8 memory corruption, V8 sandbox escape, CacheStorage race condition, and FedCM use-after-free to escape the browser sandbox.
• Safari exploit broke same-origin policy via a JavaScriptCore use-after-free, bypassed Gigacage defense, and used a sandbox IPC flaw to read history, cookies, and sessions from disk.
• Android exploit required physical access to a locked Pixel 9, using three flaws to skip USB data block, pairing authorization, and six-digit code check, then a fallback path sent raw file bytes past a hardcoded block.
• Windows remote crash bug was an out-of-bounds read in the kernel's QUIC code; a few packets could crash and restart an HTTP/3 server.
• Linux chain gave remote root access on servers with a multipath TCP listener; no login or user action needed. It first leaked a kernel pointer to defeat KASLR, then reused a race condition to overwrite a callback pointer and run a shell command.

Summarised by SecOraLens from Cyber Security News. Always read the original source before acting. How SecOraLens verifies advisories.