Debian Decides: Contributors Can Use Generative AI 'Responsibly'

Debian recently voted on whether to ban the use of generative AI tools in its projects. The outcome of the vote, however, is not to endorse or prohibit their use. The project maintains that all contributors are responsible for the technical quality, legal acceptability, and suitability of their submissions in Debian.

Generative AI is considered on par with any other tool and does not exempt contributors from the same standards of quality, correctness, maintainability, and legal compliance. It is expected that contributors will review, test, and modify AI-assisted output before integrating it into Debian. While Debian encourages disclosure of AI-assisted contributions, it is not mandatory.

The legal status of AI-generated material remains unclear, as it varies across jurisdictions. Debian does not attempt to resolve these legal questions, nor does it determine whether AI-generated content is copyrightable or derived from copyrighted works. Instead, the project relies on contributors' judgment and responsibility when using generative AI tools.

Debian's policies regarding licensing, copyright, software freedom, and contribution acceptance remain in effect, regardless of the tools used. Contributors are urged to safeguard their keys, credentials, and embargoed security reports from being leaked to third-party AI services.

The decision to ban or discourage AI use is postponed, with a recommendation to reassess the matter in two years. Some contributors express concern over the environmental impact, copyright and licensing uncertainty, the health of Free Software communities, and the strain on open web scraping. Discussion on Debian's mailing list continues, with differing opinions on locally running LLMs versus cloud LLMs and their implications for data provenance, licensing, and community health.