Pliny the Liberator, a pioneering AI red-team researcher, and John V, co-founder of the BT6 white-hat hacker collective, dive into the world of AI security. They discuss the power of universal jailbreaks, which bypass guardrails, and critique the concept of security theater. Their conversation highlights the need for open-source data, real-world safety over model restriction, and the capabilities of segmented sub-agents in orchestrating attacks. From community-driven research to the tension between sharing and misuse, Pliny and John redefine what safety in AI can be.
