Linux kernel kills AI disclosure rule after Torvalds calls it 'stupid'

The Linux kernel community has officially dropped its requirement that contributors disclose AI-assisted coding, a policy that lasted barely a year. The decision came after Linus Torvalds, the project's ultimate arbiter, called such rules 'UTTERLY STUPID' and compared them to demanding assembler code from anyone who uses a compiler.

Sasha Levin, a Linux kernel developer and Nvidia engineer, announced on the kernel mailing list that the Maintainers Summit reached consensus to deprecate the 'Assisted-by' tag and remove the documentation requiring it. The tag, introduced in January 2026 after a proposal in July 2025, had asked contributors to acknowledge when they used 'any sort of advanced coding tool' in creating a patch.

The policy was born in an era when AI-generated code was obviously buggy and bad. But as Greg Kroah-Hartman, a veteran Linux kernel maintainer, put it at the Linux Foundation's Open Source Summit in South Korea in August: 'LLMs are good enough now... they've gotten good enough to find bugs because LLMs, AI, are just pattern matchers and code is nothing but patterns in a specific format.' Torvalds had earlier said he was 'fairly positive' about vibe coding as a way for people to get into programming, but remained skeptical about its effect on software maintainability. Since then, he has clearly warmed to AI tool usage, and the policy reversal reflects that shift.

Jeff Layton, a Red Hat kernel developer, laid out three reasons for reversing course: not everyone abides by the policy; it 'basically serves as free advertising for proprietary LLM companies'; and 'it's not clear why we want to collect this info in the first place.' The policy had become, in effect, an honor system that many contributors ignored, and the tag itself was seen as a marketing boost for AI vendors rather than a meaningful governance tool.

Not everyone is happy. Laurent Pinchart, a kernel dev and founder of Ideas on Board, argued that the tag sets expectations and that unenforceability is not a reason to abandon it: 'There will be people who won't care, there will be people who cheat. We've always had those, we'll keep having them, and we'll keep rejecting their contributions.' He challenged the argument that AI labeling rules will just be ignored, insisting that setting expectations still has value, 'and I'd argue there's even more value now.'

Torvalds responded with a blistering rejection. He wrote that the argument 'has nothing to do with expectations' and 'everything to do with your hangups.' He then offered a sarcastic new rule: 'don't use compilers for any of your kernel submissions' unless you can write assembler, and don't use an editor to write commit messages because it hides mistakes. 'Do you not see how UTTERLY STUPID those kinds of rules would be?' He concluded: 'The kind of expectations I have is that people do their best and I expect people to understand what they are doing. But that understanding has very little to do with what tools they use.'

Cautious acceptance hasn't erased concerns about errors slipping through unnoticed, accountability, and coding agents reproducing copyrighted code. There's also still a stigma in some circles around the use of AI tools. Those who have spent years honing software skills may disparage developers who don't understand the code they conjure with a prompt. Rapid change will do that - in the space of a few years, software production practices have changed drastically, and software quality, security, and maintenance practices have yet to fully adjust. The leading makers of AI coding agents - Anthropic, Google, Microsoft, OpenAI, and the like - are doing their best to promote AI-driven projects as better than anything a human developer could manage. Yet even they differ in their AI attribution practices. Claude adds provenance messages by default, but Codex does not. In fields such as advertising and art, content provenance can help assess misinformation and human creative input. For code, the focus tends to be more on functionality than form.

For companies building on open source, the Linux kernel's stance could become a template for their own AI governance policies. For executives, the takeaway is that AI-assisted development is no longer a stigma in the world's most scrutinized codebase. The focus is shifting from how code is produced to whether the contributor understands it. That may accelerate adoption of AI coding agents across enterprises, but it also puts the onus on engineering leaders to verify understanding, not just output. For boards and CTOs, the signal is clear: AI tools are now part of the standard toolkit, and the competitive advantage will come from how well teams understand and verify the code they ship.