The FDA’s framework for AI regulation, while robust for premarket evaluation, would benefit from more specific mechanisms for continuous monitoring of AI performance in diverse real-world settings.