Linux kernel allows AI-generated code but developers must take full responsibility

Linux Kernel Establishes First Formal AI Policy After Months of Debate

The Linux kernel project has officially codified its first formal policy on AI-assisted code contributions, ending months of fierce debate within the open-source community 1

1

. Linus Torvalds and kernel maintainers have established clear guidelines that position AI as a development tool rather than a replacement for human developers, balancing modern AI capabilities with the kernel's rigorous quality standards 2

2

.

The new policy introduces the Assisted-by tag as a transparency mechanism, requiring developers to disclose when AI tools like Microsoft Copilot contribute to their code 4

4

. This tag serves dual purposes: enabling kernel maintainers to apply extra scrutiny to AI-assisted patches while avoiding stigmatization of the practice itself 1

1

.

Developer Accountability Anchored in New Guidelines

The policy makes developer accountability its cornerstone. AI agents cannot add the legally binding "Signed-off-by" tag—only humans can certify the Developer Certificate of Origin 3

3

. The human submitter must review all AI-generated code, ensure licensing compliance, and take full responsibility for the contribution, including any resulting bugs or security flaws 2

2

.

Code must be compatible with GPL-2.0-only licensing and include proper SPDX identifiers 4

4

. This approach legally anchors every line of AI-generated code onto the shoulders of the person submitting it, eliminating any possibility of blaming the AI for mistakes 3

3

.

Controversy That Sparked the Policy

The Assisted-by tag emerged from controversy when Nvidia engineer and prominent kernel developer Sasha Levin submitted a patch to Linux 6.15 entirely generated by AI, including the changelog and tests, without disclosing this to reviewers 1

1

. While Levin reviewed and tested the code, the lack of transparency provoked massive community backlash 2

2

. The code included a performance regression despite being reviewed, and Torvalds admitted the patch wasn't properly scrutinized partly because it wasn't labeled as AI-generated 2

2

.

At the 2025 North America Open Source Summit, Levin himself began advocating for formal AI transparency rules, proposing the first draft in July 2025 1

1

. Initial discussions on the Linux Kernel Mailing List debated whether to use Generated-by or Co-developed-by tags before settling on Assisted-by to accurately reflect AI's role as a tool 1

1

.

Pragmatism Over Panic in AI Disclosure Policy

Linus Torvalds took a characteristically blunt stance during the debate, stating he didn't want kernel documentation to take either the "sky is falling" or "it's going to revolutionize software engineering" position 1

1

. His pragmatism shaped the policy's philosophical backbone: AI is just another tool, and bad actors submitting garbage code won't read documentation anyway 2

2

.

Kernel maintainers aren't relying on AI-detection software to catch undisclosed AI-generated code. Instead, they're using deep technical expertise, pattern recognition, and traditional code review 1

1

. As Torvalds noted in 2023, "You have to have a certain amount of good taste to judge other people's code" 1

1

. He pointed out there's "zero point in talking about AI slop" because people creating it won't document their patches as such 1

1

.

Code Transparency Addresses Open-Source Community Concerns

The policy arrives as the open-source community struggles with an identity crisis over artificial intelligence 2

2

. Major projects have taken wildly different approaches: Linux distributions like Gentoo and NetBSD moved to outright ban AI-generated submissions, with NetBSD maintainers describing Large Language Model outputs as legally "tainted" due to murky copyright issues surrounding training data 2

2

.

The open-source world has been drowning in what the community calls "AI slop." The creator of cURL closed bug bounties after being flooded with hallucinated code, whiteboard tool tldraw began auto-closing external pull requests, and projects like Node.js and OCaml faced massive AI-generated patches exceeding 10,000 lines 2

2

.

AI Tools Prove Increasingly Valuable for Kernel Development

The decision reflects a shift in AI capabilities. Greg Kroah-Hartman, maintainer of the Linux stable kernel, recently noted that "something happened a month ago, and the world switched" with AI tools now producing genuine, valuable security reports rather than hallucinated nonsense 1

1

. Most AI use in kernel development is assistive—code completion, refactoring suggestions, test generation—rather than full code generation 1

1

.

The choice of Assisted-by over Generated-by was deliberate, influenced by three factors: accuracy in describing assistive rather than generative use, consistency with existing metadata tags like Reviewed-by and Tested-by, and avoiding implications that AI-assisted code is suspicious or second-class 1

1

.

As one of the most influential open-source projects, Linux's approach could establish industry standards for AI-assisted code contributions 4

4

. The policy strips emotion from the debate by acknowledging reality: developers will use AI tools to code faster, and attempting to ban specific software is impractical 2

2

. The enforcement strategy doesn't depend on catching every violation but on making consequences severe enough to discourage dishonesty 1

1

.