Policy on LLM-assisted contributions

Development
1

Hi!

A few of us work at an Italian IT company that started a small social responsibility initiative: volunteers can spend some of their work time contributing to free and open-source projects, and Syncthing is one of the projects we’d love to help with.

It’s a small thing and we won’t be able to contribute a lot, but we’d like to do it right.

Before we start, we wanted to check how you feel about LLM-assisted contributions, so we know which tools we can use. We saw the “Using AI tools to fix bugs” thread, but it’s about reporting issues, and we couldn’t find anything about code in the contribution guidelines:

  • Are you OK with code written with the help of AI tools (autocomplete, chat assistants, coding agents)?
  • If so, do you have any conditions, like mentioning AI use in the PR, having every line reviewed by a human before submitting, or only using it for some kinds of changes?
  • Same question for documentation.

If you’d rather we didn’t use these tools at all, that’s totally fine too.

Thanks!

2

The contribution guidelines says:

We do not accept contributions that are wholly or mostly AI generated.

This is somewhat vague, and in fact maybe not expressed tightly enough; it may become refined in the future. You can assume that if you feel you would need to disclose AI usage, or if it in any way looks like an AI generated change, it is not accepted. I’d be fine with someone using AI to troubleshoot, or possibly review, as long as code is manually written and the AI review is kept to yourself.

There is also

Do not submit AI generated issues or comments.

which you may have already seen.