He’s very realistic about the usefulness of AI tools for cybersecurity - his thesis appears to be that while Mythos itself is mostly PR hype, these types of models are improving at an unbelievable rate, and that in the past two months alone they’ve greatly improved in quality. He speculates that this will ultimately benefit the defender, but that right now things are changing so fast that it’s nearly impossible to predict what the future will look like.

Amazing interview, but not quite as “fuck AI” as some might hope.

https://youtu.be/1sd26pWhfmg

Related: Black Hat LLMs