I study safe human-robot interaction, particularly when robots learn from and about people. My goal is to develop robots that interact safely despite imperfect human models; for example, autonomous vehicles that automatically slow down around erratic pedestrians and assistive robots that only learn from human feedback they can understand. My research proposes novel control-theoretic frameworks for analyzing data-driven human models and develops theoretically rigorous and practical robot algorithms for safe interaction with people. A core aspect of my approach is consistently evaluating my methods through robotic hardware experiments with real human participants in domains like assistive robotic manipulators, quadrotor navigation, and autonomous vehicles.