LLMs are very good at testing for and finding exploits, especially in unfiltered models with unlimited tokens.
LLMs are very good at testing for and finding exploits, especially in unfiltered models with unlimited tokens.