AI Policy

This policy applies to all contributions — including funding proposals, milestone submissions, code, issues, reviews, comments, and documentation — to this repository, regardless of whether you are a maintainer or an external contributor.

Using AI tools is allowed and often useful. The rules below exist to protect the quality and integrity of the Development Fund process and the time of reviewers and other contributors.

You are accountable for what you submit

You are 100% responsible for every contribution you make under your name, regardless of which tools were involved in producing it. Reviewers review your work, not AI output.

You must understand and be able to defend every material element of your submission, including code, technical claims, citations, budgets, milestones, acceptance criteria, adoption claims, and security findings. You must also be able to explain why each element is included. “The AI wrote it” is not an answer.

Guard reviewers' time

Aggressively self-review AI output before asking anyone else to look at it. If you would not ask a colleague to read a specific proposal, do not ask the maintainers either.

Do not open a PR you would not have opened without AI assistance. An AI making content cheap to produce is not a reason to produce more of it.

The same applies to review comments you leave on others' PRs: do not post AI-generated feedback you have not read carefully and are not prepared to defend yourself.

Issues and PRs that appear to be unreviewed AI output — overlong and padded, referencing things that do not exist, plausible-sounding but wrong, filled with bot-style prose, or (in the case of security reports) describing a vulnerability that does not reproduce against the actual code — may be closed without further discussion. Maintainers are not obligated to explain, respond to, or debate the merits of such submissions. Repeated submissions of this kind may lead to being blocked from the repository.

If you believe a closure was in error, a single comment with a concrete, verifiable correction is welcome. Further back-and-forth is at maintainers' discretion.

Attribution

There is no requirement to mark AI-assisted contributions as such.

That said, do not hide it if asked. If a reviewer asks how a piece of code came about, answer honestly.