Linus Torvalds rejects AI-tool bans in Linux contribution dispute
His mailing-list response sets a standard for contributors: understand the work.
Loading page…
His mailing-list response sets a standard for contributors: understand the work.
Listen to this story
In an October 13, 2026 mailing-list reply, Linus Torvalds argued that banning AI or other assistance is a poor way to judge whether Linux contributors understand their work. He said contributors should use their judgment and do their best, regardless of the tools involved. The exchange began with a documentation patch to remove “Assisted-by” references, but Torvalds’s response does not mean the patch was merged or establish a new contribution policy.
Sasha Levin opened the thread on October 9 with a documentation patch, not a proposal to ban AI-generated code.
Torvalds compared excluding AI to requiring kernel developers to write in assembler or forbidding editors to test their competence.
His stated expectation is that contributors understand their work; he also praised people willing to use new tools.
Linus Torvalds rejected blanket restrictions on AI tools in an October 13, 2026 Linux kernel mailing-list reply, saying contributors should understand their work regardless of how they produce it. His intervention turned a proposal to remove “Assisted-by” documentation references into a dispute over whether tool choice is a meaningful test of competence.
The thread began on October 9 with Sasha Levin’s patch titled “[PATCH] docs: remove Assisted-by references.” Its subject was documentation, rather than a proposal to ban AI-generated code. By October 13, the discussion had reached a more basic question: what should Linux ask contributors to demonstrate about the work they submit?
Torvalds was responding to Laurent Pinchart, whose message he quoted: “What this is about is setting our expectations, telling what we want and what we don't want.” Torvalds rejected that framing and called the concept of “don't use AI for X” idiotic. His criticism targeted categorical restrictions, not simply the wording of a disclosure reference.
To make his objection concrete, Torvalds proposed a deliberately unreasonable alternative: prohibit compilers and require contributors to write kernel code in assembler. Under that hypothetical rule, using a compiler would count against a developer’s competence because it would obscure whether they understood what the machine was doing.
He then applied the same reasoning to writing a commit message, the explanation accompanying a code change. Contributors should not use an editor either, he suggested sarcastically, because deleting typing mistakes would hide evidence of their ability. Both examples challenged the idea that removing assistance is a useful way to establish understanding.
Torvalds’s response was not that contributors could stop exercising judgment. He said he expected people to do their best and understand what they were doing. He also praised people willing to use new tools. The distinction in his argument was between accepting assistance and treating assistance as evidence that the contributor lacks understanding.
But that understanding has very little to do with what tools they use.
Linus Torvalds, Linux kernel mailing list, October 13, 2026
More importantly, that reply states his position on tool restrictions; it does not announce that the documentation patch has been merged.
Loading discussion...
Join the conversation
Explain where the comparison holds up or breaks down.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.