Unreleased Claude model weaponizes Linux kernel bug into root exploit
An experimental Anthropic AI turned a known vulnerability into a working remote exploit; Ubuntu and Red Hat lack patches while Debian already shipped a fix.
What to know
- An experimental Claude model discovered how to turn a known Linux kernel vulnerability into a working remote root exploit requiring no authentication.
- Debian patched the bug on October 16; Ubuntu and Red Hat remain unpatched as of the disclosure.
- The exploit targets servers with multipath TCP listeners and allows arbitrary command execution as root.
Anthropic AI developerDebian Linux distributionUbuntu / Red Hat Linux distributions
How it unfolded 1 development · click the chart to see its coverage posts
-
1
Anthropic discloses Claude model converted kernel bug to working exploit
Anthropic publicly disclosed that an unreleased Claude model turned a known Linux kernel bug into a functional remote root exploit targeting servers running multipath TCP listeners. The exploit requires no authentication and allows arbitrary command execution as root.
“An unreleased Claude model turned a Linux kernel bug into a remote root exploit, and Ubuntu and Red Hat still lack a patch.”
— IntCyberDigest -
‼️ An unreleased Claude model turned a Linux kernel bug into a remote root exploit, and Ubuntu and Red Hat still lack a patch. Anthropic says the exploit runs commands as root on servers with a multipath TCP listener, no login needed. Debian shipped its fix on 16 October.
-
-
background
Debian ships patch for Linux kernel bug exploitable by unreleased Claude — Debian released a security patch addressing the vulnerability that the experimental Claude model later weaponized into a remote root exploit.