Linux Lays Down the Law for AI-Generated Code

The Linux kernel community has established its official stance on AI-generated contributions. In a move reflecting its pragmatic culture, the project has not banned the use of coding assistants but has issued a definitive mandate: developers bear full, undiluted responsibility for any code they submit, regardless of its origin.

The new policy, documented within the kernel's source tree, addresses the rising tide of AI-assisted patches. It makes no distinction between code written by hand, copied from a reference, or suggested by a large language model. The individual attaching their name to a patch must comprehend its function, justify its design, and defend it during the kernel's famously thorough review process. This extends the project's long-standing Developer Certificate of Origin, treating AI output no differently than any other contributed work.

Notably, the rules do not force contributors to disclose their use of AI tools. The community's position is that disclosure is secondary to competence. If a developer cannot explain their code, a disclaimer offers no protection. The guidance warns that AI is particularly ill-suited for the kernel's complex demands, such as concurrency, memory management, and hardware interaction, where superficial correctness can mask critical flaws.

The underlying concern is practical. Maintainers, a scarce resource, report an increase in patches that appear AI-generated—seemingly plausible but lacking necessary context or introducing subtle errors. The policy explicitly states these tools cannot replace human review. It's a measured counterpoint to industry hype, emphasizing that in high-stakes systems software, understanding trumps speed. The Linux kernel's approach, often a bellwether for open source, provides a clear template: use the tools if you wish, but you own the results.

Source: Webpronews