> ## Content Index
> Fetch the complete content index at: https://globalfeed.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Linux flaw rated 9.8: Anthropic says Claude wrote a remote root exploit
- URL: https://globalfeed.ai/en/linux-flaw-rated-9-8-anthropic-says-claude-wrote-a-remote-root-exploit/
- Published: 2026-10-30T23:53:05.000Z
- Updated: 2026-10-30T23:53:05.000Z
- Description: CVE-2026-98432 has fixes in the stable kernel branches, while Ubuntu and Red Hat updates are not complete. Anthropic says an unreleased model wrote the exploit, and Red Hat also lists the flaw as Critical.
- Author: GlobalFeed Editor
- Tags: Anthropic, Linux, Claude, cybersecurity, x-AnthropicAI, dil-en, elle

CVE-2026-98432, a flaw in the Linux kernel's IPv4 TCP code, received a score of 9.8 out of 10 from the kernel CVE team. [In a post dated October 27, 2026](https://www.anthropic.com/research/discovering-vulnerabilities-with-claude?ref=globalfeed.ai), Anthropic says an unreleased internal research model wrote an exploit for it after initial prompting. The exploit runs commands as root on a server with a multipath TCP (MPTCP) listener, with no credentials or user interaction. In Anthropic's proof of concept, the process took about 53 minutes.

## The patch exists, but distributions are moving at different speeds

The flaw is a race-condition use-after-free that has been in the kernel since version 4.4\. The fix went upstream and was backported to the maintained stable branches. Debian has closed it with packages 6.1.190-1 and 6.12.113-1\. According to BleepingComputer's October 30 report, fixes for Ubuntu 22.04, 24.04, 26.04 and 26.10 are in progress, and Red Hat lists RHEL 8, 9 and 10 as affected. The same report says there are no confirmed reports of exploitation in the wild.

The real fix is a kernel update. Where a package has not arrived, limiting access to trusted networks reduces the risk. CIQ recommends blocking IPv4 TCP packets that carry IP options and says disabling MPTCP alone is not enough, because the bug can also be triggered through ordinary TCP traffic. Since Go 1.24, Go programs enable MPTCP by default on TCP listeners when the kernel supports it, so administrators may be running such a listener without knowing.

## Anthropic's account and its limits

Red Hat's page rates the flaw Critical and shows a 9.8 score, while noting that the figure is preliminary and subject to review. Remote code execution on servers without an MPTCP listener has not been publicly demonstrated. Anthropic's post also covers findings that led to 62 CVEs in Chrome, along with chains in Safari, the Android lock screen and Windows QUIC. The company describes this as work with little human involvement, but it gives no success rates, costs or human hours per chain. This account rests on Anthropic's own work. For a fuller picture, follow reviews from independent researchers too.

Sources: [Anthropic](https://www.anthropic.com/research/discovering-vulnerabilities-with-claude?ref=globalfeed.ai), [Ubuntu](https://ubuntu.com/security/CVE-2026-98432?ref=globalfeed.ai), [Debian](https://security-tracker.debian.org/tracker/CVE-2026-98432?ref=globalfeed.ai), [Red Hat](https://access.redhat.com/security/cve/CVE-2026-98432?ref=globalfeed.ai), [BleepingComputer](https://www.bleepingcomputer.com/news/security/critical-linux-kernel-flaw-allows-remote-root-attacks-on-mptcp-servers/?ref=globalfeed.ai)