FDA Considers Clinician-Style Tests for Generative AI Medical Devices

Generative AI medical devices can produce different answers to open-ended prompts and may change as models or usage patterns evolve, making conventional software validation ill-suited to cover every clinical interaction. The U.S. Food and Drug Administration is therefore exploring a competency-based framework inspired by how clinicians are assessed and credentialed, a potentially important shift as AI tools move from administrative support into diagnosis, treatment and direct patient care.

In a discussion paper reported on August 18, 2026, the FDA asked stakeholders to comment on premarket testing and postmarket oversight. It proposed setting task-specific benchmarks, confirming performance in clinical settings, and comparing some devices with either a panel of qualified clinicians reflecting the standard of care or the median practicing clinician. The agency also raised the possibility of accepting greater uncertainty before authorization if stronger real-world monitoring can detect changing performance and emerging risks after deployment.