According to a report by the German technology outlet Golem, the Linux kernel project no longer requires contributors to flag code that was written with the help of AI assistants. The convention — recorded as an `Assisted-by:` trailer in a patch's commit message — has been under discussion in the kernel community for more than a year.
The trailer grew out of a straightforward question: if part of a patch was produced with a large language model, should the maintainer reviewing it be told? Supporters framed it as a matter of provenance. Kernel development has always recorded where code comes from — a sign-off line certifies that a contributor has the right to submit it, and `Co-developed-by` trailers record who did the work — so extending that record to AI assistance looked to some maintainers like the consistent move.
Others were unpersuaded. A tag is only as trustworthy as the person writing it, and there is no reliable way to inspect a patch and determine whether a machine was involved. Mandating disclosure for some contributions while the same tools sit in every editor, the sceptics argued, ends up producing a trailer that is missing precisely when someone forgets or prefers not to say — which is worse than having no convention at all.
What has not changed is the Developer's Certificate of Origin, the sign-off by which a contributor asserts the right to submit the code. The dispute was about disclosure of AI assistance, not about licensing or liability. Anyone who wants the rules in full should read the kernel's own `submitting-patches` documentation, which remains the authoritative statement of what a patch must contain; the Golem report is our source here and a second-hand summary of that policy.
