The Pulse
QGIS Bans AI-Assisted Code in Its Main Repository
QGIS merged a proposal barring AI-, LLM- or agent-assisted code contributions to its main repository. The policy also warns that violations can lead to code reversion and sanctions on contributor rights.
AI.info Team · Written with AI ·
QGIS draws a harder line on AI code
QGIS approved a ban on AI-assisted code contributions to its main repository on October 7, replacing a policy that allowed contributors to use any tools if they reviewed the generated work themselves. The new rule states: “No LLM, AI or agent assisted code contributions are allowed in the https://github.com/qgis/qgis repository.” If the project later discovers that a contributor used those tools, the policy says it may revert the code and sanction the contributor’s rights.
The decision addresses a dispute over whether human review can keep AI-assisted submissions from overloading maintainers. Supporters argue that the volume and complexity of generated code shift too much review work onto volunteers; critics have questioned how to distinguish prohibited code generation from more limited uses, such as autocomplete or translation. The final text names code contributions to the main QGIS repository, rather than banning AI across every QGIS-related activity.
From human review to a closed door
The previous QGIS policy, QEP 408, required contributors to read and review AI-generated code or text before asking maintainers to review it. It also expected contributors to disclose substantial use of generated content. The policy’s rationale was that unreviewed AI output could transfer the work of checking and correcting a contribution from its author to the project’s maintainers.
QGIS Project Steering Committee member Tim Sutton backed a clearer prohibition in a September discussion, writing:
“If that includes banning AI / LLM use in https://github.com/qgis/qgis we are fine with that.”
The resulting proposal treats a human review step as insufficient for code contributions to the main repository. Its stated penalty—code reversion and contributor-rights sanctions—makes the policy more than a request to disclose AI use.
Contributors questioned what counts as AI assistance
Discussion on the proposal raised practical questions about the boundary of the ban. Contributors asked whether it reaches LLM-backed autocomplete, AI used to brainstorm a design, or translation tools, and suggested that the policy specify code written by an LLM or agent. Those questions matter because the adopted text prohibits AI-assisted code contributions without listing permitted or excluded examples.
The debate also exposed a disagreement over how QGIS should protect its review capacity without discouraging new contributors. Supporters described a rising burden on maintainers and argued that overly broad or low-value submissions consume time needed for other work. Others warned that rules aimed at AI-generated code could be unclear in everyday development, particularly when tools mix suggestions, search and code generation.
QEP 428 leaves exceptions for later debate
The proposal, QEP 428, passed its discussion stage before moving to a vote on October 5. On October 7, its author, Nyall Dawson, declared it approved with four votes and merged it into the QGIS Enhancement Proposals repository. That merge establishes the policy in the project’s proposal repository; the text specifically names the separate main QGIS code repository as the place where the ban applies.
Dawson described the prohibition during the discussion as a starting point rather than an irrevocable rule: future proposals could argue for specific uses, such as AI autocomplete or translation, one case at a time. For now, the adopted text contains no such exceptions. It says that AI-, LLM- or agent-assisted code contributions are not allowed in the main repository, and warns that discovered use can lead to code reversion and sanctions on contributor rights.