Finally, we talk about how Anthropic and OpenAI agreed to give each other temporary access to each others LLMs for safety/security testing, which is somewhat unheard of and fascinating as an evolution of safety testing.
Lots of fun conversation and good to be back at it!