Research audio from Failure-First: adversarial evaluation of embodied AI, jailbreak archaeology, policy analysis, and daily paper summaries from the AI safety frontier.
