Linux Kernel Officially Opens the Door to AI-Generated Code

Linux Kernel Writes the Rulebook for AI Code

The Linux kernel, the open-source foundation underpinning everything from Android phones to cloud servers, has formally codified rules for AI-assisted contributions. The new documentation, now live in the kernel's official repository, permits developers to use AI coding tools while placing full legal and technical responsibility squarely on the human submitter.

The policy was authored by stable kernel maintainer Sasha Levin of Nvidia and backed by Linus Torvalds himself. It emerged from consensus reached at the 2025 kernel maintainers summit in Tokyo and was merged into the kernel repository this past week.

Human Accountability at the Core

Under the new framework, AI agents are explicitly barred from adding Signed-off-by tags, the legal mechanism through which contributors certify compliance with the kernel's GPL-2.0 licence. Only a human can sign the Developer Certificate of Origin. The submitter must review all AI-generated code, ensure licensing compliance, and take complete ownership of the contribution.

Contributors are required to include a new Assisted-by attribution tag specifying the AI tool name, model version, and any specialised analysis tools used.

A Policy for Good Actors

Torvalds and Levin have made clear the policy targets good-faith contributors, acknowledging that bad actors would simply not disclose AI usage regardless of any rule. The approach focuses on transparency and human responsibility rather than attempting to police or detect AI-generated submissions.

The move comes as AI tools have become deeply embedded in developer workflows, with surveys showing eighty-four percent of developers now using AI coding tools. Kernel maintainers have already integrated an AI code review system called Sashiko into their workflow.

Setting a Template for Open Source

The Linux kernel's approach may serve as a blueprint for other open-source projects navigating the same questions. The policy sidesteps the contentious debate over whether AI-generated code trained on copyrighted material can be properly licensed, instead focusing on the practical question of accountability. Community discussions have flagged unresolved copyright concerns, but the kernel team has opted for pragmatism over perfection.