One of the open-source world's oldest and most influential projects has begun formally voting on how to handle contributions made with large language models. Debian developers opened a General Resolution vote on August 15 that could ban, restrict or fully embrace AI-assisted work across the distribution — a decision being watched closely by other free software projects wrestling with the same question.
Voting runs until August 28, with the ballot listing eight competing proposals plus a "None of the above" option, as reported by Linuxiac. Only Debian Developers may vote, using the project's ranked-choice voting system to order their preferences. For latest AI developments touching the free software community, the vote is the culmination of a debate that has run for months — and, in different forms, for more than two years.
A debate years in the making
Debian's struggle over AI-generated code is not new. In May 2024, the project dismissed an early effort to establish an AI-contributions policy, judging the question premature. A General Resolution on AI models and the Debian Free Software Guidelines was withdrawn in May 2025 before a vote could conclude. This summer the debate reignited, with developers arguing on mailing lists through July over whether LLM-assisted work belongs in a project whose Social Contract promises users a completely free operating system.
Now the question has finally reached the ballot — with an unusually wide spread of answers.
The case for a ban
The original proposal would amend the Debian Social Contract itself to ban LLM-assisted contributions: no generative AI for direct Debian work such as packaging, software, documentation, translations, websites or project communication. Upstream projects made with AI help would not be affected.
The proposal's stated rationales read like a summary of the free software movement's objections to generative AI: unclear copyright and licensing status of model outputs, doubts about the reliability of AI-generated work, extra review burden pushed onto maintainers, aggressive scraping of Free Software resources for training data, and the enormous resource consumption of large AI systems.
At the strictest end, another proposal asks contributors to avoid LLM use as much as possible, requires that all Debian communication — bug reports, mailing-list messages, Salsa discussions, even Planet Debian blog posts — be written only by humans, and would enforce violations through the project's Code of Conduct.
The middle ground: allow it, but own it
Between the poles, several proposals would permit AI-assisted contributions while placing responsibility squarely on the human who submits them.
One option would clearly allow contributions made partly or fully by an LLM, provided contributors verify their technical quality, security, licensing and usefulness, fully understand what they are submitting, and disclose significant AI help. It would also prohibit sending private or sensitive Debian information to untrusted external AI services.
A similar proposal acknowledges the concerns around generative AI but argues a ban would be unenforceable and counterproductive, accepting AI-assisted work as long as it complies with the Debian Free Software Guidelines, is properly reviewed and understood by the contributor, and is marked as AI-assisted where necessary.
The most hands-off option says Debian should neither support nor ban generative AI tools, instead applying the same standards of quality, correctness, maintainability and legality to every contribution regardless of how it was made. Contributors would still have to review, understand, test and take full responsibility for AI-assisted work, with disclosure encouraged but not required.
A gentler variant would encourage contributors to prefer human authorship and understanding without imposing a project-wide ban, keeping disclosure voluntary.
"Debian is created by humans"
Among the most discussed options is a proposal titled, simply, "Debian is created by humans." Rather than policing what tools contributors use, it focuses on what enters the project: contributors could use generative AI for research, analysis, exploration or critique, but could not submit AI-generated output directly as Debian packaging, patches, documentation, bug reports, project communications or other official work. The philosophy is captured in the title — humans, not models, make Debian.
The eighth and final proposal takes a different lens entirely, focusing on the environmental impact of large-scale AI systems and what Debian's response to it should be.
Why a Linux distribution's vote matters
Debian is far more than one distribution. It is the foundation for dozens of derivatives including Ubuntu, which in turn underpins everything from cloud infrastructure to Linux-based developer tooling worldwide. Policies adopted by Debian developers tend to ripple outward — into derivative distributions, into packaging practices, and into the broader norms of free software development.
The vote is also a test case for a question every software organization now faces: when code, documentation and bug reports can all be generated by a model, what does it mean to contribute? Some projects have already tightened contribution policies around AI-generated content, while others have embraced it wholesale. Debian's ranked-choice ballot — with eight distinct answers on offer — is the most granular expression of that dilemma any major open-source project has produced so far.
Whichever option prevails on August 28, the outcome will set a precedent: either the first major distribution to formally restrict LLM-assisted work, or a carefully-worded framework for how one of the internet's foundational software communities decides to live with the machines trained on its own output.
Stay Ahead of AI Read more AI news →
