If you have screenshots or other proof I'd try to contact their security team directly. https://openai.com/.well-known/security.txt But might also end up with auto-reply emails to use their bug tracking system

I had an issue where one completely unrelated conversation about a topic bled into one that happened days ago. When I asked about it, ChatGPT said:

> It looks like I accidentally analyzed and described a Hammerspoon script instead of pulling data from your science benchmark PDFs—sorry about that!

Very weird.

I remember seeing something similar happen around the time OpenAI launched ChatGPT. I can't seem to find it right now, but at one point they served users' entire chat histories to the wrong users, even in different languages.

undefined