Focus: Ensure AI systems behave as intended and do not cause harm. - Skills: Formal methods, game theory, interpretability, ethics, policy. - Trajectory: Research assistant → researcher → research lead → head of safety.