Post-Cutoff

Policy & safetyLinux FoundationIn its training data

Torvalds: the ‘continued flood of AI reports’ has made the Linux kernel security list ‘almost entirely unmanageable’

Confirmed

The takeaway

In his Linux 7.1-rc4 announcement (May 17, 2026), Linus Torvalds wrote that AI bug-finding tools had flooded the kernel’s private security list with duplicate reports.

Status
Claim

Confirmed

Our reporting
High confidence
Importance
3 of 5
Last verified
2 October 2026

Your AI and this story

  • GPT-6.2 AstraIn its training data
  • Claude Opus 5.5In its training data
  • Gemini 4 Hydrogen47 days after its cutoff
  • Grok 4.7In its training data

It is after the cutoff of Gemini 4 Hydrogen, and can be in the training data of GPT-6.2 Astra, Claude Opus 5.5 and Grok 4.7.

Key facts

  • Torvalds (LKML, 7.1-rc4): ‘the continued flood of AI reports has basically made the security list almost entirely unmanageable, with enormous duplication due to different people finding the same things with the same tools’
  • Also: ‘People spend all their time just forwarding things to the right people or saying “that was already fixed a week/month ago”’. AI-found bugs are generally not secret, because the same tools are available to everyone. ‘If you found a bug using AI tools, the chances are somebody else found it too’
  • July 2026: the kernel CNA published 432 CVEs over a single Sunday and Monday; press linked the surge partly to AI-assisted vulnerability research (Techzine, Open Source For You)
  • Security Boulevard (Sept 2026): ~500 CVEs per release in the 6.x era, 1,000+ for Linux 7.0, 1,500+ for 7.2 and approaching 2,000 for 7.3. It attributes the rise to AI tools including Anthropic’s Mythos and OpenAI’s GPT-5.6 Sol scanning old driver code
  • Sept 29, 2026: Debian DSA-6528-1 (kernel 6.12.111-1 for Debian 13) fixed 1,000+ CVEs (9to5Linux counts 1,313). The advisory itself does not mention AI; 9to5Linux attributes the size to the kernel’s assign-a-CVE-to-every-security-fix policy and Debian batching several upstream releases
  • Caveat: since February 2024 the kernel is its own CVE Numbering Authority and assigns CVEs liberally, so CVE counts overstate the rise in serious vulnerabilities

What happened

After frontier models with strong vulnerability-finding skills spread in spring 2026 (Claude Mythos Preview / Project Glasswing, OpenAI’s Daybreak), many researchers pointed the same AI tools at the Linux kernel and reported the same bugs to its private security list. Torvalds said the list had become almost unmanageable and asked people to add value with fixes. Kernel CVE output kept rising through the summer, and downstream distributions shipped record-size kernel advisories.

Why it matters

This is a concrete case of AI bug-finding at scale shifting the bottleneck from discovery to triage and patching for volunteer maintainers. It also shows the duplication problem: when everyone runs the same models, AI-found bugs are effectively public.

Sources

7 sources from 7 sites. Numbers match the chips in the text.

7 sources: 2 primary, 5 press

Primary

  1. LKML: Linus Torvalds, Linux 7.1-rc4 announcementlkml.org, official
  2. LWN: Debian DSA-6528-1 kernel security updatelwn.net, docs

Press

  1. The Register: Torvalds says AI-powered bug hunters have made Linux security mailing list ‘almost entirely unmanageable’theregister.com, press
  2. Techzine: 432 Linux vulnerabilities put pressure on patch managementtechzine.eu, press
  3. Open Source For You: Linux maintainers battle record AI-fuelled CVE surgeopensourceforu.com, press
  4. Security Boulevard: AI helps drive number of Linux kernel CVEs to near 2,000 per releasesecurityboulevard.com, press
  5. 9to5Linux: Debian 13 kernel security update patches more than 1300 CVEs9to5linux.com, press

Changes

  • Filed (the advisory itself does not attribute the CVEs to AI)

Status

Claim

Confirmed

Our reporting
High confidence
Importance
3 of 5
Last verified
2 October 2026

Sources at a glance

7 sources: 2 primary, 5 press

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
2 October 2026
Human review
None recorded for this entry. What the editor does
Version
Last saved 4 October 2026

Spotted an error? Write to contact@postcutoff.com. Corrections are logged in public.

This page for your AI

Same text, no layout:

Open in ClaudeOpen in ChatGPT

Related

Related events

  1. Policy & safety

    OpenAI launches Daybreak cyber-defense initiative with GPT-5.5-Cyber and Codex Security

    Confirmed

  2. Model releases

    Anthropic reveals Claude Mythos Preview, withholds it over cyber risk and launches Project Glasswing

    Confirmed

The whole topic