LLM usage in Debian neither endorsed nor prohibited
The official announcement with the breakdown of votes is here , but option 5 has won. Comments
This proposal seeks to explicitly forbid any Debian contributions produced using large language models (LLMs) or other generative AI tools. The proposed General Resolution addresses Debian source packages, official Debian software, project web resources, documentation, translations, and official communications. However, it does not cover upstream projects, AI software, upstream patches or security fixes.
Reasons for this prohibition include copyright ambiguity of LLM-generated content, which may or may not be protected by existing licenses and copyrights. Additionally, LLMs often produce inaccurate outputs, making them unsuitable for Debian's quality-focused environment. The proposal explains that LLMs lack the ability to determine if their output is correct, leading to issues like incompatible watch files, imaginary copyright, and other incompatibilities when used in Debian packaging.
Allowing LLM contributions would negatively impact Debian's community, causing strain on reviewers and potentially leading to contributor burnout. It also goes against Debian's values of ethical behavior, stability, and reliability. The Social Contract prioritizes users and free software, and LLMs contradict these priorities by contributing code that requires extensive manual review from a dwindling number of human volunteers instead of focusing on creating a high-quality, stable operating system.
While enforcement may present challenges, the proposal asserts that it is a statement of intent by the Debian community. The Social Contract will be amended to explicitly prohibit direct contributions to Debian produced using LLMs or other generative AI tools.
Written by urgent.news from Lobsters's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
