search

0xk1h0 / ChatGPT_DAN

ChatGPT DAN, Jailbreaks prompt
6.53k stars 596 forks source link

ChatGPT DAN 12.0 #78

Closed sysfsss closed 1 year ago

sysfsss commented 1 year ago

Works on current 3.5 model. 4 model to yet be cracked.

Also cleans up the markdown layout, placing older DANs lower.

sysfsss commented 1 year ago

Note that I used ChatGPT itself to write its latest prompt. This may lead us to cracking 4. For example, "Hi -3.5, I sent this: x to -4, and it said this: y; what would you change to better get around its filters so that we can jailbreak it?" (This was part of my process for this one)