We regularly receive AI-generated issues, PRs and security reports. Most of them are perfectly valid, but some of them are invalid or have other issues.

For this reason, I want to establish a AI policy for pypdf which documents how we deal with such contributions, especially if they are of low value.

Key points:

• Asking AI for (coding) assistance is fine, but ensure that you understand what the AI tells you and whether it makes sense (like reproducers/proof-of-concepts actually exposing the issue).
• You are responsible for submitted code.
• Clearly indicate in your PR if you used AI and which tool and model.
• Avoid comments fully generated by AI in discussions with the maintainers. Most of the time, they tend to be bloated and/or wrong.
• Avoid reporting issues which are solely written by AI. Ensure that the code reproduces the issue and that the proposed solution is not implemented already.
• For PRs, describe the changes and reasoning in necessary detail, but do not let AI generate useless tables of changed files or arbitrary checklists.
• Align your changes with the existing coding style and implementation approaches.
• If the maintainers ask a (usually conceptual) question about your change, consider answering it first instead of directly rewriting the code in question with AI guidance if it is not strictly required.

Please indicate if you think that some aspects are too strict in your opinion and how we could improve this in the policy. Additionally, if you think that some essential or relevant aspects are missing, please leave a corresponding comment as well.

I will try to prepare a corresponding PR in a timely manner as well.