Policy proposal: AI-assisted contributions

[vissarion]

Due to a recent increase in PRs and issues that appear to be AI-assisted, I think we should add a clear generative-AI policy to our contribution guidelines.

The goal isn’t to ban AI, but to require transparency and keep review effort reasonable by making contributors explicitly disclose AI usage (tool/model + version, where used, and the scope of assistance) and affirm that humans reviewed, validated, and take full responsibility for the result.

Several communities already follow this “allow, but disclose and stay accountable” approach.

One very good example is https://joss.readthedocs.io/en/latest/submitting.html#ai-usage-policy

Some ideas:

• Generative AI is welcome for most contributions (code/tests/docs/PR text), but any use must be disclosed in the PR (tool/model + version, where used, and what it did).

• Human contributors and reviewers must remain fully responsible: humans must review/validate all AI-assisted output, make core design decisions, and personally verify all technical claims.

• AI should not be used to “speak for” authors/reviewers in discussions (except translation with disclosure)

• Not following the policy can lead to PR rejection or correction actions.

• New contributors should not open large PRs

• New contributors should not open more than one (parallel) PR