Linus Torvalds on Linux and AI: "Fork It or Walk Away"

Linus Torvalds does not do subtle.

The creator of Linux posted a message on the Linux Kernel Mailing List this week that left no room for interpretation. Linux is not an anti-AI project. Developers who object to AI tools being used in kernel development are free to fork the project or leave it entirely. And anyone who tries to talk other contributors out of using AI will be loudly ignored.

For a community that has spent years debating whether AI-generated code belongs anywhere near one of the world's most critical software projects, this was about as clear a statement as anyone could have asked for.

Here is what Torvalds actually said, what triggered it, how his views have shifted since 2024, and what it means for the future of Linux development.

What Torvalds Actually Said

The statement came in response to a thread on the Linux Kernel Mailing List that had grown heated around the use of AI tools in kernel maintenance workflows.

Torvalds wrote directly:

"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. Or just walk away."

He continued: "AI is a tool, just like other tools we use. And it's clearly a useful one. It may not have been that clearly even just a year ago, but it's no longer in question today. Anybody who doubts that clearly hasn't actually used it."

On the question of forcing anyone to use AI, he was equally clear: "We're not forcing anybody to use it, but I will very loudly ignore people who try to argue against other people from using it."

His position as top-level maintainer of the Linux kernel gives this statement real weight. He was not offering an opinion for debate. He was setting policy.

What Triggered This Statement

The immediate flashpoint was a tool called Sashiko — an AI-powered patch review system built specifically for the Linux kernel by Google engineer Roman Gushchin. Sashiko operates under the Linux Foundation and automatically scans submissions to the Linux kernel, flagging potential issues before human maintainers review them.

A post from the Software Freedom Conservancy titled "Recommendations When Using LLM-backed Generative AI Systems for FOSS Contributions" had been circulating in the thread. Some developers read it as a signal that AI tools should be treated with significant caution or discouraged. Others were more directly opposed to Sashiko's expanding role in the review process.

The debate had been simmering for a while. Sashiko's presence in the kernel workflow had raised questions about whether individual developers should be able to opt out of receiving AI-generated review comments, and whether automated AI analysis should carry any formal weight in the patch acceptance process.

Torvalds had seen enough. His response made the kernel's official position on this clear.

"The growth of AI coding tools has not been limited to the kernel. OpenAI recently launched the Codex Micro keyboard, a physical hardware device built entirely around AI-assisted coding workflows."

How His View Has Changed Since 2024

What makes this statement particularly notable is how different it sounds from what Torvalds was saying less than two years ago.

In October 2024, Torvalds described nearly 90 percent of AI as "marketing hype" and said he intended to largely ignore it. He predicted that the technology would mature over the following five years — implying it was not yet ready to be taken seriously as a core development tool.

By May 2026, his tone had shifted. Speaking at the Open Source Summit North America in Minneapolis, he compared AI's impact on programming to the introduction of compilers. The technology was changing how software gets written, he said, but not the fundamentals of what programming actually requires. A developer still needs to deeply understand the generated code and the system architecture — knowing how to write a good prompt is not the same as knowing what the code should do.

Now, in July 2026, he has gone further. He is not just acknowledging that AI is useful — he is actively defending developers' right to use it and telling opponents that their objections will not shape the project's direction.

His own words capture the shift best: "It may not have been that clearly even just a year ago, but it's no longer in question today."

Where He Still Draws the Line

Torvalds is not saying Linux should be built by AI. He is not advocating for AI-generated patches to bypass the review process. And he is not saying AI is without problems.

He has been direct about the downsides too. AI can increase maintainer workload when it surfaces large volumes of unvetted reports that still require human investigation. He acknowledged that AI is "a somewhat painful tool, both for maintainer workloads and just from a it keeps finding embarrassing bugs standpoint."

On the question of over-reliance, his position is even firmer. He has said that when he sees people claiming that 99 percent of their code is written by AI, "I literally get angry." The concern is not that AI writes code. The concern is that developers stop understanding what the code does.

His view is essentially this: AI is a legitimate tool in the engineering workflow. It is judged by the same standard as every other tool — does it produce technically useful output? But the humans using it are still responsible for the result, and understanding the code you submit remains non-negotiable.

What Is Sashiko and Why Does It Matter?

Sashiko is the AI system at the center of this debate and it is worth understanding what it actually does.

Built by Roman Gushchin, a kernel engineer at Google, Sashiko performs agentic review of Linux kernel code changes. It automatically scans patches submitted to the kernel, analyzes them for potential issues, and generates review comments — all before a human maintainer sees the submission.

The Linux Foundation has been running it as part of the kernel's maintenance infrastructure. For maintainers dealing with an enormous volume of patch submissions, automated pre-screening that can flag obvious problems before human review is a genuine time-saver.

The concerns raised by some developers were not entirely without basis. An automated system that generates review comments at scale has real risks. If the signal-to-noise ratio is poor — if Sashiko flags too many non-issues — it adds work rather than reducing it. Torvalds himself has previously complained that AI-powered bug-hunting reports had made the Linux security mailing list difficult to manage.

But his response to those concerns is not to ban the tool. It is to hold the output to the same standard as everything else: does it actually help, measured by technical merit?

What This Means for Linux Development

Torvalds' statement does several things at once.

It ends the possibility of a blanket anti-AI rule being adopted at the kernel level. Individual contributors can still choose not to use AI tools themselves. They cannot veto colleagues who do.

It signals that AI-assisted review systems like Sashiko have a legitimate place in the kernel workflow, provided they produce technically useful output. The standard is not ideology — it is whether the tool helps maintainers do their jobs.

It also puts the Linux kernel in a different position from some other open source projects. Other communities have taken stricter stances, with at least one major open source project rejecting any AI-generated code on the grounds that it raises unresolved copyright questions.

Torvalds acknowledged that there are real open questions around AI — particularly around the economics of the technology and its long-term impact. But usefulness is no longer one of those open questions, in his view.

The practical outcome of all this is not dramatic. AI-generated patches do not get a faster path into the kernel. Human review, subsystem ownership, testing, and regression analysis remain exactly as they were. What changes is that the kernel community now has a clear maintainer-level policy: AI use is permissible when it produces technically useful work, and principled objections to the technology itself will not carry the day.

"AI is already automating repetitive development tasks in other ways too — tools like Codex Record and Replay let developers demonstrate a workflow once and have AI repeat it automatically."

The Broader Significance

Linux sits beneath a significant portion of the world's computing infrastructure. The kernel runs on servers, smartphones, cloud platforms, embedded systems, and developer workstations across every industry. Decisions made about how the kernel is developed have long-term consequences that extend well beyond the open source community.

Torvalds taking a clear pro-tools stance on AI — not pro-AI in the ideological sense, but in the practical sense that useful tools get used — sends a signal to the broader software development world.

The debate about AI in software development is not going away. Questions about code quality, copyright, maintainer workload, and the appropriate role of automated systems in critical infrastructure are all live issues that the industry is still working through.

What Torvalds has done is take one specific question off the table for the Linux kernel: whether AI tools are categorically disqualified from participating in the process. They are not. Everything else — which tools, how they are used, what standards their output is held to — remains to be worked out in practice.

Models like Claude Sonnet 5 have been built specifically for agentic coding work — exactly the kind of AI assistance Torvalds says developers should be free to use.

Frequently Asked Questions