How should FDA regulate a medical AI system that can generate a different answer every time?
FDA is exploring competency assessment, risk-based monitoring and new lifecycle controls for generative AI medical devices whose outputs can vary by prompt, context and software evolution.