Open Source Project QGIS Bans All AI Code to Stop Maintainer Burnout
QGIS just banned AI-generated code. Maintainers are drowning in low-quality pull requests and losing their best reviewers.
Policy · Source: Hacker News
What happened
QGIS is a massive open-source geographic information system trusted by organizations worldwide. They just slammed the door shut on AI. A new policy strictly bans all AI-assisted code contributions across the project. The pull request, named QEP428, merged with overwhelming support from core maintainers. This is a direct response to an unsustainable burden placed on the people who keep the project alive. Free software communities are built on human networks, and AI is currently breaking those networks.
The old policy required a human in the loop. It failed completely. Maintainers got flooded with AI-generated pull requests. These submissions looked plausible on the surface but contained subtle, complex bugs. Reviewing them took too much time. It broke the fundamental trust contract between contributors and reviewers. Contributors used AI to fix bugs they did not actually understand. A few tokens later, the bug looked fixed, but the contributor learned nothing about the codebase.
The burnout is real and measurable. The project already lost at least one extremely high-value contributor to the stress of reviewing AI code. Maintainers call this ban a necessary stop-gap measure. They are closing the door completely right now to stop the bleeding. They might inch it open later for specific uses like IDE autocomplete or translations, but only after careful discussion. Two years ago, the world functioned without AI. Maintainers want to return to that baseline until the AI bubble settles.
Key facts
- QEP428 — The official enhancement proposal banning AI contributions
- ~50 — Number of pull requests tagged as AI Assisted under the old policy
- ~40 — Number of those AI Assisted pull requests that were closed without merging
Why it matters
Open source software runs entirely on human trust and limited bandwidth. When you submit code to a major project, reviewers assume you understand the architecture. LLMs destroy this assumption. They give beginners the illusion of competence. They generate code that compiles but lacks deep system context. Maintainers are forced to act as debuggers for AI output instead of peer reviewers for human logic. This shifts the entire cost of development onto the reviewer. The person generating the code pays nothing. The person reviewing it pays with their time and sanity.
This ban is a massive signal. It is the first major backlash against AI coding tools in established open-source communities. It will not be the last. Expect more large projects to follow the QGIS model. If AI tools make writing code frictionless but reviewing code harder, the open-source ecosystem will collapse under the weight of unreviewed pull requests. Projects will have to build automated AI reviewers or ban AI entirely to survive. The economics of open source are changing. Writing code is now cheap. Trust and review are the new premium assets.
For builders
Stop spamming maintainers with AI code
Maintainers will ban you or your tool. If you use AI to write a pull request, you must understand every single line. You pay the price in reputation if you waste a reviewer's time with generated garbage.
Build AI tools for code review
The bottleneck is no longer writing code. It is reviewing it. Founders who build AI agents that can accurately audit and verify AI-generated pull requests will solve a massive pain point. Open source projects and enterprises will pay for this.
My take
I build AI products for a living, but I completely agree with the QGIS maintainers here. We gave junior developers a machine gun that shoots pull requests, and we pointed it directly at unpaid open-source reviewers. Until AI can review code as well as it writes it, strict bans are the only way these communities survive.
Original reporting: Hacker News. This is my rewrite and opinion.