AI safety or jailbreaking

on

 

How Johnny Can Persuade LLMs to Jailbreak Them:
Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs