Repository navigation
Should we have an AI policy? #649
Description
What is the Topic of Your Question
Other
Add Your Question Below
I'm just thinking a section in the CONTRIBUTING.md about whether AI assisted code is allowed (I assume yes) and whether agent-only PRs are allowed (I assume no???).
We can link to the ARC policy when it's published. But just wondering about the vibe for this particular project.
@paddyroddy @dstansby @matt-graham Any thoughts? Should I make a proposal?
Activity
dstansby commentedon Oct 7, 2026
hello from the past, my opinion shouldn't really carry any weight now. My two cents is no AI (co)authorship, because you should take responsibility for your code even if AI wrote it, PRs should be reviewable by humans (including keep them well focused and as short as possible), AI code is fine.
paddyroddy commentedon Oct 9, 2026
I think it's probably a good idea to have such a policy in this modern world. Surprised this didn't get flagged in the ARC work recently, i.e. glass-dev/glass#1185. Agree with @dstansby.
pradeeppeddineni commentedon Oct 25, 2026
I'd say yes go for the proposal, the replies here already agree on most of it. Something this short would cover it:
That's close to what LLVM does (human in the loop, label substantial tool output, no agents acting without approval: https://llvm.org/docs/AIToolPolicy.html) and the kernel, where only a human can add Signed-off-by (https://docs.kernel.org/process/coding-assistants.html).
One thing to decide is if this is only for contributions to this repo, or if the template should also put a section in generated projects. The {{cookiecutter.project_slug}} folder has no CONTRIBUTING.md right now so that's probably a separate issue tbh.