跳到正文
原文
Decrypt· Jose Antonio Lanz·· 2 小时前精选AI 评分70

Claude 智能体发现 Ubuntu 漏洞,Ubuntu 尚未修复

Claude-Built Linux Exploit Targets Bug That Ubuntu Still Hasn't Patched

AI 导读

Anthropic 的 Claude 智能体发现了一个 Linux 漏洞,该漏洞可能被用于远程控制 Linux 服务器。尽管 Red Hat 已发布修复程序,但 Ubuntu 仍存在漏洞。

推荐理由

原文揭示了 Claude 智能体发现的一个 Linux 漏洞,该漏洞可能对 Ubuntu 用户构成安全风险,读者可以了解其影响和应对措施。

原文

In brief

  • Anthropic says Claude wrote an attack that gives a stranger full control of a Linux server over the internet, using a bug that has existed since Linux version 4.4.
  • Red Hat released fixes on Sunday, Nov. 1, but Ubuntu still lists every version from 16.04 to 26.10 as vulnerable.
  • The Go Security Team says the best fix is to upgrade to a patched kernel, and warns the bug is not limited to the feature Anthropic's attack used.

Another day, another exploit from ChatGPT or Claude that gives us the chills.

Anthropic says its AI wrote a working attack that lets a stranger take full control of a Linux server over the internet, no password needed. The bug behind it, CVE-2026-98432, has been hiding in Linux since version 4.4 and scores 9.8 out of 10 for severity.

Linux runs countless servers, and Ubuntu is one of its most popular flavors. Ubuntu's own tracker still lists every version from 16.04 to 26.10 as vulnerable.

$94.16K volume

Closes: Dec 31, 2026, 23:59 UTC

Closes: Dec 31, 2026, 23:59 UTC

Anthropic described the attack on Oct. 27 in a research post covering five targets: Chrome, Safari, Android, Windows, and Linux. Claude wrote the Linux exploit after initial prompting, using an unreleased internal model.

The attack takes less than an hour to find a hidden spot in the computer's memory, then five more minutes to run a command with the highest level of access, known as root.

The flaw works like reading a sticky note after someone already shredded it. One part of the system throws away a piece of information while another part is still using it, per AlmaLinux. A specially crafted message that opens an internet connection can set that off.

Failing costs the attacker almost nothing. "An enabling property of this race condition is that a lost race only fails the handshake without crashing the whole system," Anthropic wrote, "so a determined attacker can retry indefinitely until they succeed."

Anthropic's version needs a server using a feature called multipath TCP, which lets one connection travel over several routes at once. Programs written in Go have switched it on by default since Go 1.24, per the Go Security Team.

But the bug is not limited to that feature. "While the demonstrated exploit requires the target system to expose an MPTCP listener, the underlying bug is not specific to MPTCP," the Go Security Team wrote.

Red Hat released fixes for RHEL 8, 9 and 10 on Sunday, including RHSA-2026:93412. Debian, SUSE, Oracle Linux and Amazon Linux have also shipped fixes, at least for some versions.

For anyone who can't update yet, Red Hat published a temporary firewall rule. It "reduces exposure to this vulnerability but cannot fully prevent it," Red Hat says in its security bulletin. A full working solution is still under development.

Anthropic's earlier Claude Mythos model helped Mozilla find 271 bugs in Firefox, as Decrypt reported in April.

Ubuntu's tracker shows fixes in progress for 22.04, 24.04, 26.04 and 26.10. Versions 20.04, 18.04 and 16.04 show only "Vulnerable," and the tracker gives no release date for any fix.

Daily Debrief Newsletter

Start every day with the top news stories right now, plus original features, a podcast, videos and more.

Join 72,000+ readers

Create an account to save your articles.

Add on Google

Add Decrypt as your preferred source to see more of our stories on Google.

来源:Decrypt · decrypt.co