If you wanted to know why LLMs red-teaming is not really red-teaming, why fuzzing is unlikely to solve all the issues with LLMs, why LLMs are so data-hungry, what are limitations of prompt engineering and how this all relates to cyber-security, my keynote from Insomnihack (@1ns0mn1h4ck) is up on Youtube: https://youtu.be/kJ1Kh3RxczQ

Once again, thanks to SCRT and @1ns0mn1h4ck for giving me an opportunity to talk about my research and the general state of the field!

#LLM #cybersecurity