Linux kernel maintainers have taken a notable step in their handling of AI-assisted programming. The Register reports that the community has agreed to stop asking contributors to disclose whether AI tools helped produce their patches. The change follows a consensus at the Maintainers Summit to deprecate the "Assisted-by" tag, and Sasha Levin, a kernel developer and Nvidia engineer, relayed it to the kernel mailing list last week with a request to remove the documentation that called for the tag.

Torvalds' stance

The Register highlights that Linus Torvalds has said plainly that he opposes "stupid" AI disclosure rules, provided people understand the code they submit. The principle is straightforward. The kernel has long relied on a chain of human responsibility, where each patch has an identifiable author who answers for it. In that view, the tool is secondary. A contributor who cannot explain or defend a change is a problem whether the code came from a chatbot, a snippet copied from elsewhere, or a tired late-night edit.

A shift from earlier friction

The article frames the decision against earlier controversies over AI in the kernel community. The new consensus suggests a cultural softening, with the maintainers who set norms for the project becoming more comfortable with AI assistance. The Assisted-by tag had been a middle path, allowing AI use but asking for transparency. Deprecating it indicates the maintainers concluded that the tag added little beyond bureaucracy.

Why it matters for AI coding tools

The kernel is one of the most visible and demanding software projects in the world. Its code runs on servers, phones, embedded devices and supercomputers, and its review process is known for being strict. When such a project drops mandatory AI disclosure, it sends a signal to other open source maintainers and to companies drafting internal rules. The message is that AI-assisted code is acceptable when the human submitter stays fully accountable.

The excerpt does not name any specific AI coding products, so the story does not directly shift the competitive standing of particular assistants. The likely effect is indirect. Less social stigma and less paperwork around AI use may encourage more kernel contributors to use assistants in their workflow, which could benefit the broader category of coding tools.

Trade-offs and open questions

Disclosure tags offered several benefits. They told reviewers when extra scrutiny might be warranted, they created a record for any later questions about code provenance or licensing, and they gave researchers data on how AI-generated code behaves in a mature codebase. Without them, the project depends on reviewers' instincts and contributors' candor. If AI lowers the cost of producing plausible but subtly flawed patches, maintainers could see a heavier review load. Other open source communities have raised related worries about low-quality AI-generated submissions, and The Register has run a related piece on the topic.

The approach also leaves questions unanswered. Kernel contributors still must certify that they have the right to submit their code under the project's licensing rules, and how AI-generated material fits that certification is a legal gray area that a tag deprecation does not resolve. The practical answer under Torvalds' framing is that the submitter owns that responsibility.

Industry context

Levin's employer, Nvidia, is a major participant in kernel development and in AI infrastructure. His role in announcing the outcome illustrates how closely corporate engineers are involved in shaping kernel norms. Hardware and cloud vendors have strong incentives to see AI-assisted development become frictionless, though the decision was reached by the maintainers as a group and reflects Torvalds' long-standing preference for judging code on its merits.

Bottom line

The kernel community has moved from asking how a patch was made to asking whether the author understands it. The shift reflects the growing normalization of AI in professional software development. It places the burden of quality on contributors and reviewers instead of on labels, and it will be watched by other open source projects deciding how to handle AI-assisted contributions. How well the policy holds up will depend on whether review capacity can absorb any increase in AI-generated submissions.