Conceivably ChatGPT could help, with more suggestions for fuzzing that independently operating malicious actors may not have been able to synthesize.

Most of the really bad actors have skills approximately at or below those displayed by GPT-4.