The Linux kernel community is retiring its requirement that contributors identify patches created with advanced coding tools, moving its AI policy away from mandatory disclosure and toward individual responsibility for submitted code, The Register reported on October 14.

Kernel developer Sasha Levin said the Linux Maintainers Summit reached consensus to deprecate the “Assisted-by” tag and remove the documentation that required it. The rule had instructed developers to add the tag to a commit message when advanced coding tools helped create a patch.

Code patch moving through abstract review, bug, licensing, and maintenance checks.
Maintainers see useful roles for coding agents while retaining concerns about errors, accountability, and copied code.

The disclosure proposal first emerged in July 2025 and entered the kernel documentation in January 2026, according to The Register. Since then, AI coding agents have improved enough that some maintainers now consider them capable of producing acceptable work and helping locate bugs, even as concerns remain about errors, accountability, and the reproduction of copyrighted code.

Supporters of removing the tag questioned whether it produced useful information. Red Hat kernel developer Jeff Layton argued in an earlier discussion that compliance was inconsistent, that the tag effectively promoted proprietary model vendors, and that the project had not established a clear purpose for collecting the data.

Fading disclosure tag beside a chain connecting a developer, code patch, and system core.
The new direction places responsibility on contributors to understand the code they submit, whatever tools they use.

Not every contributor agreed. Kernel developer Laurent Pinchart argued that a disclosure rule could still communicate the community’s expectations even if some people ignored it. He warned that removing the tag without another mechanism could make it harder to ask submitters about AI involvement.

Linux creator Linus Torvalds rejected the idea that tool declarations were a meaningful test of competence. His position, as reported by The Register, is that contributors should understand the work they submit and do their best, regardless of whether they used an AI system or another development tool.

The decision does not amount to an endorsement of unreviewed generated code. It changes what contributors must reveal, not who remains responsible for a patch. For the kernel community, the emerging standard is therefore practical rather than provenance-based: code must be understood, reviewable, and maintainable, while the particular tools used to produce it become less central to the submission record.