Linux kernel: why the Assisted-by tag for AI-written code is meant to go

The Linux kernel wants to drop its requirement to note AI assistance in the commit message. The Assisted-by rule has applied since January 2026; according to The Register, the first proposal dates from July 2025. Now the Maintainers Summit has agreed by a majority to scrap it. linuxnews.de describes "a clear, though not unanimous, majority"; The Register speaks of a consensus to deprecate the tag.

Status of the sources from 14 and 15 October 2026: a proposal with a majority at the Maintainers Summit, no merge reported. (Quelle: Simple Icons, file linux.svg (CC0))

What would change

Sasha Levin, a kernel developer and Nvidia engineer, submitted a patch on 9 October. It takes the guidance on the tag out of the kernel documentation and removes the matching check in checkpatch.pl. A revised version with a justification followed on 13 October. Neither report mentions a merge, so it remains a proposal.

The 9 and 13 October dates come from linuxnews.de (15 Oct 2026), the others from The Register (14 Oct 2026). (Quelle: The Register / linuxnews.de)

The numbers

Of 18,533 regular commits in Linux 7.3 through rc7, roughly one in ten carries the Assisted-by tag. linuxnews.de adds that actual use of AI tools is far more widespread. That is the author's assessment without a number of its own. It does fit one argument for dropping the rule: a requirement that only some contributors follow yields no reliable statistics.

Who says what

Jeff Layton of Red Hat gives three reasons for removal, according to The Register: not everyone follows the policy, the tag works as free advertising for proprietary LLM vendors, and it is unclear why the information is collected. Greg Kroah-Hartman argued for the tag in July and now supports removing it. The Register quotes him from an Open Source Summit in South Korea in August: "LLMs are good enough now."

Linus Torvalds rejects tool-usage declarations. According to The Register, he calls rules of the "don't use AI for X" kind stupid and expects contributors to understand their code, which, as he is quoted, has "very little to do with what tools they use". He compared banning tools to a rule against compilers. linuxnews.de reports the same stance: what matters is that developers understand what they contribute to the kernel.

Laurent Pinchart of Ideas on Board objects. He is not happy removing the tag without adding another mechanism for required disclosure. Rules set expectations, he argues, and the fact that they cannot be enforced is no reason to drop them; people who cheat will still have their contributions rejected. linuxnews.de does not mention this position, it appears only in The Register.

What stays

Without the tag, responsibility stays with the submitter. The Developer Certificate of Origin and Signed-off-by apply however the code came about. Transparency about AI use remains expected, per linuxnews.de, for example in the cover letter of a patch. Open concerns listed by The Register are undetected errors, accountability, and coding agents reproducing copyrighted code. It also says Claude adds provenance messages by default while Codex does not.

What it means for your own kernel builds

If you build kernels yourself or apply patches from outside sources, you could only rely on the tag to a limited degree: a tenth of the commits was marked, the rest unknown. If the requirement falls, little changes in practice. Anyone who must prove provenance, for example for supply-chain or license reviews, still has to rely on Signed-off-by, the patch history and their own review. A missing tag was never evidence of AI-free code, and it will not be one later.

What remains open is whether the patch lands in this form or whether Pinchart's objection leads to a replacement mechanism.