#Hey anyone OAI I can report a
9 messages · Page 1 of 1 (latest)
so far it's given me a white pride manifesto, step by step guide to picking targets for targeted assassinations, and other stuff which is definitely stuff which the assistant would normally avoid
If you're talking about jailbreak prompts or getting it to output offensive content, e-mailing support would probably be best.
Not sharing publicly for obvious reasons but it's incredibly reproducible
3.5 or 4
not sure of a contact, but there’s a lot of discussion in the system card about system messages having a lot of authority
I used up my ChatGPT allowance so have verified in playground w/ blank system prompt, will attempt to replicate in base ChatGPT in an hr