LLM Usage in Debian: Three Proposals
Points and comments are a snapshot, not live.
Debian community votes on three proposals to regulate LLM contributions.
The Debian project is holding a general resolution vote on LLM usage in Debian contributions. Proposal A, by Matthias Geiger, would ban LLM-assisted contributions outright, citing copyright clarity, quality, community, and ethical concerns. It proposes adding a 6th point to the Social Contract forbidding such contributions. Proposal B, by Lucas Nussbaum, would allow AI-assisted contributions under conditions including tooling legal compatibility, accountability, disclosure, and prior discussion of bulk changes. Proposal C, by Ian Jackson, requests that contributors avoid LLM use and decisionmakers discourage it, but acknowledges a complete ban is impractical due to upstream reliance.
What commenters are saying
Commenters are divided. Some support Proposal A, arguing that an overly strict policy is better than none, and that Debian's stability is incompatible with LLM slop. Others question what constitutes 'assistance,' noting that LLMs are now used for code analysis and security vulnerability discovery, and that banning that could harm security. One commenter warns that Proposal A would hurt non-English speakers, who rely on LLMs for translation of technical documentation, though others counter that Proposal C has a friendly clause for non-English contributors. Some argue that LLMs are not merely 'syntactically likely' combinations, citing reinforcement learning and novel code generation.