north dock
Hey there! If you think you've found a security vulnerability, we have a bug bounty program or you can contact the security team directly. Details here: https://chat.openai.com/.well-known/security.txt
If you're not sure, maybe you can give some more details on why you think it's leaking other people's responses?
candid relic
Just curious: what is the string you're generating responses to -- the thing that starts with /9j/4AA... and is really long?
I see! I'm speculating, but I'm wondering if maybe it is getting confused trying to contextualize the text of the base64 as text, rather than as an image. This could explain the random responses you're getting? Almost like going "shuffle mode" with the model?
north dock
@spare wyvern Could you DM me a link to the shared chat (even if it doesn't show the last part of the conversation)?