QGIS has adopted a ban on AI-assisted code contributions to its main repository. Nyall Dawson’s policy proposal merged on October 7, replacing the project’s earlier requirement for human review of AI output.
The merged change says, “No LLM, AI or agent assisted code contributions are allowed” in the QGIS repository. If the project later discovers such use, the policy provides for reverting code and sanctions on contributor rights.
Human review no longer grants permission
The previous policy let contributors choose their tools if they reviewed generated code or text, understood their work and could answer reviewers’ questions. It required disclosure of substantial tool-generated content, with an Assisted-by commit trailer as one example. It also barred agents from posting without human approval.
The new text removes that conditional permission. A contributor’s own review and disclosure no longer satisfy the rule for AI-assisted code.
In the discussion, GitHub user haubourg described submission volume exceeding maintainers’ review capacity. Dawson said the earlier process had failed in practice and that the project had already lost at least one highly valued contributor as a result.
Exceptions remain a matter for later proposals
Dawson said the intent was to begin with a strict ban, then consider individual exceptions through discussion. He named autocomplete and translation as possible future cases.
Participants questioned how broadly “assisted” should apply, including AI-backed autocomplete and brainstorming. GitHub user timlinux described compliance in terms of whether contributors wrote the code themselves in good conscience. Those comments leave a distinction between the broad policy text and how individual uses might be judged.
