What's the difference between AI safety and AI ethics?

They overlap but emphasize different things. AI ethics is broadly about whether a system's use is fair, transparent, accountable, and respects rights — think bias, privacy, consent, and societal impact, often here-and-now and value-laden. AI safety focuses on building systems that behave reliably and as intended, avoiding harmful failures — robustness, avoiding unintended behaviors, and (for advanced systems) alignment and control. Safety includes near-term concerns like a medical model failing on edge cases and longer-term concerns about powerful, autonomous systems. In practice teams need both: an unsafe system can cause harm even with good intentions, and a safe-but-unethical system can work exactly as designed toward a harmful goal. Many practitioners treat safety as one pillar within responsible AI.