Debian is putting the question of AI-generated code to a formal vote. According to Hacker News, a group of Debian developers has filed a General Resolution titled “LLM usage in Debian,” opening a discussion period on 2026-07-24 with three competing proposals on the table. The lead proposal, backed by roughly nine developers including Matthias Geiger and Ian Jackson, would flatly forbid any contribution to Debian written with the help of large language models.
This matters because Debian isn’t some hobby project. It’s the base that Ubuntu and much of the server world are built on. A binding rule here would be one of the strongest institutional stances against generative AI in open source to date.
What’s being proposed
The most aggressive option, listed as Choice 1, wants to amend Debian’s Social Contract itself. It would add a new clause stating the project “will not allow direct contributions to Debian written with the use or assistance of large language models (LLMs) or other generative AI tools.”
The scope is deliberately specific. Per Hacker News, the ban would cover:
- Debian source packages
- Official project software like lintian
- Debian web resources
- Documentation and translations from contributors
- Official Debian communication
It would not touch upstream projects that use LLMs, AI-related software itself, or upstream patches and security fixes. So the line is drawn at what Debian contributors themselves produce, not what the rest of the world ships.
The case for a ban
The proposal authors lay out four arguments, and they’re worth understanding because they reflect concerns far beyond Debian.
- Copyright. LLM output has murky legal status. Debian Policy demands absolute clarity on licensing, and the authors argue AI code can’t meet that bar.
- Quality. They point out that a model “can never know if its output is correct,” and that generated packages tend to mix outdated syntax, broken watch files, and “imaginary copyright” that a newcomer wouldn’t know how to fix.
- Community. Reviewing AI slop burns out the volunteer maintainers who keep Debian alive, and contributors leaning on LLMs never actually learn the craft.
- Ethics. This one is pointed. The authors say AI scraper bots have hit Debian’s own infrastructure like “a large scale and perpetual Denial of Service attack,” forcing sites offline and pushing the project to add JavaScript-based verification checks.
What stands out here is the framing. The authors tie the whole thing back to Debian’s identity, arguing that “move fast and break things” is the opposite of what makes Debian reliable.
Why this is significant
Until now, most open-source projects have handled AI contributions with informal guidance or case-by-case judgment. A few, like Gentoo, have already banned AI-generated content. Debian formalizing a position through a General Resolution, the project’s highest decision-making mechanism, carries real weight. If it passes, it becomes constitutional-level policy, not a maintainer’s preference.
The existence of three separate proposals tells you the developer base isn’t unified. A full Social Contract amendment is one path. The other choices, filed by different developers, signal that plenty of contributors want something softer, like disclosure requirements or narrower limits, rather than an outright prohibition.
One practical wrinkle worth flagging: enforcement. Detecting LLM-assisted work is genuinely hard, and the proposal leans on contributor honesty more than any technical gate. That’s a tension the vote won’t fully resolve.
What to watch next
Debian’s process runs on a Condorcet vote after the discussion period, so developers will rank all the options rather than pick one. Expect the debate on the mailing lists to get loud, because this touches identity as much as policy.
A few things to keep an eye on if you work in or near open source:
- Precedent. Whatever Debian decides, other large projects will cite it.
- The disclosure middle ground. The softer proposals may prove more durable than a total ban.
- Infrastructure pressure. The scraper-bot complaint is real and shared. That fight isn’t going away regardless of how the vote lands.
The voting outcome will say a lot about how the free software world plans to coexist with generative AI. Full details on all three proposals are available at the original source.