News · 2026-07-15

Torvalds: 'Linux is not one of those anti-AI projects'

Linus Torvalds has told the Linux kernel community that the project will not take a stand against AI. "Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it," he wrote in a mailing list message on July 14. The remark, made in passing in a thread about kernel tooling, is the clearest signal yet on where the world's most important open-source project sits in a fight that has split the rest of the ecosystem.

Key facts

• The quote: "Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it." -- Linus Torvalds.
• When: Posted Tuesday, July 14, 2026, at 20:06 Pacific, in a thread titled "Re: Linking Patchwork with Sashiko?" on the linux-media mailing list.
• Who: Linus Torvalds, creator and lead maintainer of the Linux kernel.
• Primary source: The message on lore.kernel.org; reported by Phoronix.

The context you need

Over the past year, a real schism has opened in open source over AI-assisted code. Some projects now ban AI-generated contributions outright. Some require disclosure. Some have gone further and treat using an AI tool as a mark against the contributor personally. The Flathub app store banned AI slop and watched much of its catalog vanish. The creator of the Zig language called the AI-rewrite narrative marketing. Maintainers, who are volunteers drowning in review load, have watched a flood of confident, plausible, subtly wrong patches arrive from people who did not read them.

Their frustration is legitimate. The question Torvalds was answering is whether the response should be a project-level position against the tools.

His answer, per Phoronix's report of the exchange, is no -- and the reasoning is characteristically unsentimental. He rejects what he describes as social-warrior attitudes toward developers who use AI, and he does not want the kernel to become a venue for that argument. AI is a tool. The kernel judges patches, not the provenance of the keystrokes.

The fork line at the end is not a throwaway. It is Torvalds invoking the actual constitutional structure of open source: this is a project with a maintainer, the maintainer sets direction, and the remedy for disagreeing with direction is not to win an argument about ideology but to take the code and go. It is the same freedom that makes Apache-licensed releases like xAI's Grok Build meaningful even when the maintainers won't take your patches. Fork is the ballot.

Why it matters

Two reasons, one obvious and one less so.

The obvious one is precedent. Linux is not a project among projects; it runs essentially all of the internet, every Android phone, and the fleets these AI models are trained on. When its maintainer declines to adopt an anti-AI position, every smaller project debating one now has to explain why it is stricter than the kernel. That does not settle the argument, but it moves the burden.

The less obvious one is what the position actually is, because it is easy to misread as an endorsement. Torvalds is not saying AI-generated patches are good. He is saying the kernel already has a mechanism for that question, and it is called review. The kernel's standards -- does it compile, does it not break userspace, does the maintainer understand it, is the provenance signed off -- are agnostic about how the code came to exist. A patch written by an AI and not understood by its submitter fails on the "not understood" clause, which was already a rule. Adding a separate anti-AI rule on top would be redundant at best, and at worst it would move the project's attention from code to people. That is the part he seems most keen to avoid.

The honest caveat

This is a mailing list remark in a thread about something else, not a policy document. Torvalds has been sharply critical of AI hype before, and nothing here retracts that -- "not anti-AI" is a long way from "pro-AI." The kernel has not announced guidance on AI-assisted contributions, has not changed its Developer Certificate of Origin, and has not addressed the harder unresolved questions: what a submitter is attesting to when AI wrote the patch, how copyright attaches, or what maintainers do about volume. The tools that produce the patches are getting better fast -- Anthropic now merges more Claude-written code than human-written code into its own products -- and the kernel's review capacity is not scaling with them. Torvalds has ruled out one answer. The real one is still open.

Primary source, verified: read the paper →

Ground Truth. (2026, July 15). Torvalds: 'Linux is not one of those anti-AI projects'. Ground Truth. https://groundtruth.day/news/torvalds-linux-is-not-an-anti-ai-project.html

@misc{groundtruth:torvalds-linux-is-not-an-anti-ai-project,
  title  = {Torvalds: 'Linux is not one of those anti-AI projects'},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {jul},
  url    = {https://groundtruth.day/news/torvalds-linux-is-not-an-anti-ai-project.html}
}

Topics: open-source · linux · ai-coding · community · linus-torvalds