Linux kernel allows AI-assisted code, as long as you follow these rules

The use of AI-powered tooling is becoming increasingly common in most development environments. Notable examples in this area include GitHub Copilot, Anthropic Claude, ChatGPT Codex, and more. As such, it's also natural that big organizations will begin to integrate AI-assisted code in their enterprise software too. For example, Microsoft says that AI copilots are responsible for hundreds of thousands of pull requests per month in its internal code. Now, we have some more clarity around the status of AI code in the Linux kernel too.

A relatively recent document explaining the requirements for AI-assisted code has been making the rounds on Hacker News. This RST document is available in the Linux repository owned by head honcho Linus Torvalds on GitHub. It provides guidance for developers leveraging AI-assisted code for contributions to the Linux kernel.

For starters, any AI-assisted code should follow the same processes as those followed by humans, outlined in development-process.rst, coding-style.rst, submitting-patches.rst. In addition, all code must comply with the Linux kernel's existing compliance requirements, such as compatibility with GPL-2.0-only, leverage appropriate SPDX license identifiers, and conform to all the rules outlined in license-rules.rst.

Moreover, AI agents are not allowed to use the Signed-off-by tags, as they are reserved for human reviewers only. Only humans can legally certify the Developer Certificate of Origin (DCO), and the submitter must be responsible for reviewing all AI-generated code, following licensing requirements, using their signed-off-by tag, and taking sole ownership of their contribution.

Proper attribution for AI-assisted code is also required as it helps track the evolution of this metric in the Linux kernel. Pull requests should use the Assissted-by tag in the following convention: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]. This nomenclature is described in more detail below:

• AGENT_NAME is the name of the AI tool or framework
• MODEL_VERSION is the specific model version used
• [TOOL1] [TOOL2] are optional specialized analysis tools used (e.g., coccinelle, sparse, smatch, clang-tidy)

As developers and organizations continue to embrace AI and its benefits in coding activities, it is not surprising to see the Linux kernel welcoming contributions that leverage this technology too. We already know that Torvalds himself has been utilizing vibe coding practices in some of his projects too.