Debian has launched a project-wide vote on how generative AI should be used in its open-source ecosystem, with proposals ranging from a complete ban to regulated acceptance that could shape AI governance across open source.
The Debian Project has launched a General Resolution (GR) to determine its long-term policy on the use of large language models (LLMs) and generative AI in project contributions, opening what could become one of the most influential governance debates in the open-source ecosystem.
Four competing proposals are under consideration, ranging from a complete prohibition on AI-assisted contributions to regulated acceptance with mandatory disclosure and contributor accountability. The discussion began on 24 July 2026, followed by at least two weeks of debate before Debian developers cast their votes through the project’s preferential voting system.
Proposal A seeks to ban LLM-generated contributions across Debian packages, documentation, translations, project software, websites and official communications. Supporters cite hallucinations, outdated coding practices, licensing uncertainty, ethical concerns and reduced learning opportunities for new contributors. The proposal would also add the ban to the Debian Social Contract.
Proposal B, submitted by former Debian Project Leader Lucas Nussbaum, permits AI-assisted contributions under six conditions, including legal compatibility, licence verification, contributor responsibility and disclosure using Git trailers such as Generated-By: or Assisted-By:.
Two compromise proposals discourage unrestricted AI use while recognising a total ban may be impractical. They require disclosure, allow teams to prohibit AI-generated work, mandate compliance with the Debian Free Software Guidelines (DFSG), and prohibit cloud AI services for sensitive Debian content.
The debate follows years of discussion within Debian and reflects a broader movement across projects including Gentoo, GNOME, Codeberg and FreeBSD to define governance around AI-assisted open-source development.
