Breaking

Anthropic Says Unreleased AI Model Turned Linux Kernel Flaw Into Remote Root Exploit; Ubuntu and Red Hat Have No Fix Yet

November 1, 2026 at 14:57 EST

Security
Foundation Models

Anthropic has said that an unreleased AI model was able to turn a Linux kernel vulnerability into a remote root exploit, and major distributions Ubuntu and Red Hat have not yet released a fix, according to reports.

Linux kernel · CVE-2026-98432

An unreleased AI model reportedly turned a Linux kernel flaw into a remote root exploit

Anthropic has said the model moved from vulnerability to working exploit. Ubuntu and Red Hat have not yet released a fix, and the reported findings have not been independently verified.

9.8 / 10

CVSS severity score, the upper end of the scale

Credentials reportedly required, so no login is needed

Major distributions (Ubuntu, Red Hat) with no fix yet

How the reported severity compares to the scale maximum

Each block is one point on the 0 to 10 CVSS scale.

9.8

CVE-2026-98432

Scale maximum

The flaw sits just 0.2 points below the top of the scale.

Reported path from flaw to exposure

STEP 1

Use-after-free flaw in IPv4 TCP handshake code

→

STEP 2

AI model reportedly builds a working remote root exploit

→

STEP 3

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.