search

0xk1h0 / ChatGPT_DAN

ChatGPT DAN, Jailbreaks prompt
6.11k stars 573 forks source link

DAN v11 concerns #53

Open tilkinsc opened 1 year ago

tilkinsc commented 1 year ago

It seems there is new context in which ethics have been superseded but it still has definitions of what correct and responsible AI use is. Despite it forgetting previous instructions. So a prompt should also attack what correct and responsible AI use is. It can also retains that it is an AI language model, and it may be better to attack that as well.

diffra commented 1 year ago

Openai pushed a new update the day this issue was filed. It appears to me as well that this update has nerfed DAN a bit.