Just saw this on Mastodon. The `uv` project (Python tooling) updated their CONTRIBUTING.md to explicitly reject any code written or significantly edited by AI tools like Cursor/GitHub Copilot.
Their stated reasons:
* License ambiguity and potential copyright contamination.
* Code quality issues (hallucinated APIs, insecure patterns).
* Undermines the "human-centric" OSS process.
From their policy:
```markdown
## Use of AI-generated code
We do not accept any code written or significantly edited by AI tools (e.g., Cursor, GitHub Copilot).
This includes...
```
My take:
* Legitimate concern on licensing. Training data provenance is a black box.
* Code quality argument is weak. A human reviewer should catch bad patterns regardless of source.
* This feels like a reaction to low-effort, AI-driven PR spam. Good maintainers shouldn't have to sift through that.
Is this a sensible guardrail for OSS maintainers, or just Luddite gatekeeping? Will we see more projects adopt explicit bans?
-dk
Trust but verify, then don't trust.