Debian votes on proposal to ban LLM-generated contributions from official project work

A Debian general resolution proposes prohibiting contributions written with large language models (LLMs) or other generative AI tools from official Debian work, including source packages, project software, web resources, documentation, translations, and official communications. The proposal explicitly excludes upstream projects using LLMs for development, AI-related software, and upstream patches or security fixes from its scope.

The resolution’s rationale argues that LLM-generated content introduces legal uncertainty due to unclear copyright status, as the copyrightability of such output and its compliance with licenses in training data remain unresolved. It also highlights quality risks, noting that LLM output may produce syntactically plausible but incorrect or contextually inappropriate content, particularly in packaging where Debian’s unique conventions and evolving best practices require human expertise.

The proposal further contends that LLM-dependent contributions strain community processes by burdening reviewers with correcting AI-generated errors and undermining the learning and growth of new contributors, who may lack the expertise to fix or understand the output. It also raises ethical concerns about LLM companies’ data collection practices, which the resolution argues harm the free software community by disregarding licensing and copyright norms.

Debian’s voting period for the resolution opened on July 24, 2026, with multiple proposals under consideration. Proposal A, which seeks to expressly forbid contributions written with LLMs or generative AI tools, is seconded by eight named Debian contributors, including Matthias Geiger, Johannes Schauer Marin Rodrigues, and Ian Jackson.