Earning trust in autonomous driving requires more than building state-of-the-art models—it demands a safety focus that spans hardware, software and comprehensive validation as a whole. In this talk Chen Wu presents Waymo’s approach to AI safety, from the on-vehicle sensing stack (LiDAR, vision, radar and audio) through core autonomy capabilities including the onboard system and large-scale simulation. She will explain how Waymo evaluates safety holistically using complementary roles: the “Driver” that operates in the real world, the “Simulator” that generates and replays diverse conditions and the “Critic” that measures performance against safety requirements. Chen will also cover Waymo’s Safety Framework, a layered safety determination approach. Attendees will leave with a concrete framework for thinking about AI safety for autonomous driving beyond the models—what to measure, where failures surface and how to validate at scale.