AI assistance when contributing to the Linux kernel

What the new guidance says

The Linux kernel project has published a short but decisive guide for using AI coding assistants when contributing code. The document, added to Documentation/process/coding-assistants.rst in the kernel tree, tells developers that AI tools must obey the existing development process — from coding style to patch submission — and that all contributions remain subject to the project's licensing rules. In plain terms: AI can help, but it can't skip the rules.

Licensing, DCO, and human responsibility

Crucially, the guidance repeats what many legal teams have been warning about: all kernel code must be GPL-2.0-only and use appropriate SPDX identifiers. AI agents MUST NOT add Signed-off-by tags — only a human can certify the Developer Certificate of Origin. The human submitter is explicitly responsible for reviewing any AI-generated code, ensuring license compliance, adding their own Signed-off-by, and taking full responsibility for the contribution. This is where the rubber meets the road: accountability remains human.

Attribution and traceability

To improve traceability, the project asks contributors to include an Assisted-by tag when AI tools are used, in the format Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]. Optional analysis tools (coccinelle, sparse, smatch, clang-tidy) may be listed; basic utilities like git or gcc should not. Example in the doc: Assisted-by: Claude:claude-3-opus coccinelle sparse. Simple, but important — a breadcrumb trail for how a patch was produced.

Why this matters

This is a pragmatic compromise: acknowledge AI's role without letting it erode legal and procedural safeguards. The kernel is the bedrock of countless systems; trust and provenance aren't optional. Will this calm the licensing debates around AI-generated code? Time will tell. For now, the message is clear — AI can be a tool, but the human still signs the dotted line.

Sources: github.com/torvalds, Hacker News