Btw, LLMs are already used in vulnerability discovery and exploit development.
Which you should've done before making such statements imo.