Debian voted to accept AI-written contributions, with contributors responsible

Debian's voting developers adopted a policy on generative AI. Stevie Bonifield reported it for The Verge on 31 August. Proposals that would have banned AI-assisted contributions were among the options considered and were not adopted.

What the policy says

The core sentence: generative AI "is neither exempt from nor subject to special rules beyond the standards already expected of Debian contributors."

The policy acknowledges that responsible use can improve productivity. It does not require contributors to disclose when they used AI, though it encourages disclosure.

The most important paragraph, quoted from the policy:

The Debian Project nevertheless expects that all contributions submitted to Debian, regardless of how and with which tools they were produced, satisfy the same standards of quality, correctness, maintainability, and legal compliance. The use of a generative AI tool does not diminish the contributor's responsibility for the work they submit. Contributors are expected to understand, review, test, and, where appropriate, modify AI-assisted output before incorporating it into Debian. Blindly accepting or uploading AI-generated material without appropriate human review is inconsistent with Debian's established development practices.

Why this is the right approach

Read what it does not do. It sets no percentage limit, requires no label, and creates no separate review path for generated code.

It attaches one obligation to the contributor: understand what you submit. Everything else follows from rules that already existed.

That approach avoids the mistake most AI policies make. A disclosure requirement produces a checkbox nobody verifies. A ban produces contributors who do not mention it. An obligation to understand the code is checkable in the only place it can be checked, which is when someone asks the contributor a question about their own patch and finds out whether they can answer.

Not everyone agrees

The Verge cites It's FOSS reporting that some contributors are unhappy, including one who says he is leaving and is "no longer interested in anything coming from Debian." Canonical faced similar reaction earlier this year over Ubuntu's position, and Linus Torvalds has said Linux is "not one of those anti-AI projects."

That disagreement is real and it will keep happening. Debian's answer is at least a coherent one that a maintainer can apply to a specific patch, which is more than most policies manage.

What to copy

If your organisation has not written this down, Debian's version is a good starting draft and it is short enough to adapt in an afternoon.

Same quality standards regardless of tool. The person submitting is responsible. Understand, review, test and modify before it goes in. Disclosure encouraged, not required.

The last one is the part people argue about, and Debian's reasoning is implicit: a disclosure rule you cannot enforce is a rule that only catches honest people.

Source

• Debian won't ban AI code from its Linux distribution, The Verge, 31 August 2026