Anthropic says an unreleased Claude model built five exploit chains, including a one-click Chrome attack with 62 CVEs patched in Chrome 156

Anthropic's Frontier Red Team published "Discovering and exploiting vulnerabilities in critical, real-world software with Claude" on October 27. The account below comes from a post summarizing the article, not from the article itself.

• What it describes: an unreleased internal Claude model built five exploit chains: a one-click remote code execution in Chrome, a Safari same-origin-policy bypass, an Android lock-screen bypass on the Pixel 9, a remote denial of service in the Windows kernel's QUIC code (CVE-2026-33096) and a zero-click root exploit in Linux MPTCP (CVE-2026-98432).
• Chrome: the 62 Chrome CVEs are all rated critical or high and are patched in Chrome 156. The post's advice is to update the browser.
• Status: Anthropic says all of the flaws are patched.
• Recommendation: Anthropic recommends scanning code at pull-request time, using a public Claude Code security plugin.