The New Rules for AI-Assisted Code in the Linux Kernel: Essential Insights for Developers

Introduction

In a significant move for the development community, Linus Torvalds and the Linux kernel maintainers have unveiled the project's first formal policy regarding AI-assisted code contributions. This policy not only reflects a balance between leveraging emerging AI technologies and upholding rigorous quality standards but also sets clear expectations for developers contributing to the Linux kernel.

Key Principles of the AI Policy

The newly codified guidelines are underpinned by three core principles aimed at ensuring accountability and transparency within the kernel development process:

Only Humans Can Certify Code

Firstly, it is emphasized that only human developers are permitted to add "Signed-off-by" tags to contributions. This requirement ties into the Developer Certificate of Origin (DCO), which affirms the legal compliance of code contributions. Consequently, even if a patch is generated by AI, the submitting developer must take full responsibility for it.

Mandatory Attribution for AI Tools

Secondly, the guidelines introduce a requirement for contributors to include an "Assisted-by" tag for any code that utilizes AI tools. This tag serves as a means of transparency, clearly identifying the AI models and tools employed in the development process. For instance, a correct attribution may look like: "Assisted-by: Claude:claude-3-opus coccinelle sparse."

Full Human Liability

Finally, the policy establishes that developers are wholly accountable for reviewing any AI-generated code to ensure compliance with licensing regulations, as well as for rectifying any potential bugs or security issues that may arise. This principle acts as a deterrent against attempts to submit faulty or malicious code, reminiscent of the incident involving University of Minnesota students in 2021, where they submitted flawed code that was not properly disclosed.

Addressing Controversies Around AI Contributions

The decision to formalize the Assisted-by tag was heavily influenced by previous controversies in the Linux community. Notably, Nvidia engineer Sasha Levin faced backlash after submitting a patch for Linux 6.15 that was entirely AI-generated, without disclosing the use of AI to reviewers. This incident highlighted the necessity for rules governing AI's role in kernel contributions, which culminated in the recent adoption of the AI policy.

The Need for Transparency

At the 2025 North America Open Source Summit, Levin advocated for increased transparency concerning AI-generated contributions, proposing what would ultimately evolve into the current policy. The term "Assisted-by" was selected to clarify AI's function as a coding tool rather than a co-author, emphasizing the human-centric nature of Linux kernel development.

The Role of AI in Kernel Development

Despite the inclusion of these guidelines, maintainers are cautious about relying on AI-detection software to unearth undisclosed AI-generated patches. Instead, they continue to lean on established techniques such as deep technical expertise and rigorous code review processes to ensure quality control.

The Importance of Code Quality

Torvalds has voiced concerns regarding the influx of poor-quality AI-generated code, stressing the importance of judging the value of contributions based on traditional review practices. As he has articulated, the key issue lies not with recognizing obviously flawed patches, but rather with identifying credible submissions that may inadvertently embed subtle yet critical flaws.

Conclusion

The adoption of this new policy marks a milestone in how the Linux community approaches the integration of AI tools into software development. It reconciles the innovative possibilities offered by AI with the longstanding commitment to code integrity and developer accountability. By mandating human oversight and transparent attribution, the Linux kernel aims to maintain its reputation for quality as it embraces the future of coding technology.