Anthropic AI Finds Critical Linux Kernel Flaw Giving Root Access

What was affected

The problem sits in the kernel, the core of Linux that handles basic functions such as memory and device management. Anthropic demonstrated the attack against servers that accept Multipath TCP connections, a mechanism that uses several network paths at once.

By abusing a fault in how traffic is processed, the AI gained administrator-level control with no login. The victim did not need to click a suspicious link or take any other action.

Red Hat rates the flaw 9.8 out of 10 on its severity scale, placing it in the most severe category. Linux is not common on everyday desktop PCs, but it is widely used in servers, network equipment, and industrial and research systems, so the potential reach is broad.

What admins should do

An attacker with administrator rights could steal or alter stored data, including personal information. A fix only reaches a server once the operator applies the update distributed by its Linux provider.

Amazon Linux, for example, has already published an update addressing the vulnerability. Anyone running servers should check whether their version is affected and whether the necessary update has been installed.