[I’m not sure where to ask it] — Can someone share details about the privacy implications of the “AI summarize” feature? Specifically, is the PR code stored elsewhere now? can it be used to train chatGPT for other companies?

We emailed openAPI to opt out of future training using our data per this on their website. cc @Greg who has the latest status there.

got it. thx.

Hi @micha! I appreciate the care for data privacy. • Like Aryaman said, we’ve opted out of training sets • OpenAI’s ToU says that all input is owned by the requester, not them • We’ve also crafted the feature to be PR-by-PR opt in. If you dont press the button, the diff wont get processed Does this line up with your expectations?

Yes, pretty much. Part of my [previous] hesitations around Graphite were around privacy, but I really like your privacy policy and feel more comfortable now. With chatGPT and sending potentially proprietary data elsewhere, that’s another risk. You may want to add an option to disable this for the entire org in the future, though I agree that it’s not urgent as users can just choose not to use it

Its a good callout, I appreciate the feedback. We can investigate org-settings to disable

If you want to read further or want to share our privacy expectations around the use of OpenAI with others in your org, we've added a section for it in our docs here: https://graphite.dev/docs/privacy-and-security#new-ai-summarize