FDA Plans New Policy for Generative AI Medical Devices

The agency will shift from binary software testing to evaluating clinical judgment as it prepares new guidance.

Updated on Oct. 20, 2026 in Artificial Intelligence

Bold flat-color editorial illustration of a geometric medical aperture, symbolizing new regulatory frameworks for AI-driven clinical judgment. — The FDA is developing new regulatory guidelines for generative AI in medical devices, shifting from static code testing to evaluating clinical competency. AI Illustration. Upload story photo >

Live Poll

Should government agencies regulate AI medical tools like human doctors rather than traditional software?

The FDA plans to draft new policy guidance for regulating generative AI medical devices following a public comment period that closed on October 19, 2026. The agency intends to move away from traditional binary software testing toward a model that evaluates clinical judgment and safety boundaries.

Why it matters

Traditional software validation methods are insufficient for generative AI tools, which operate in an unbounded evaluation space. The agency is now prioritizing a framework that treats these systems more like the clinical competency of human physicians.

The agency is transitioning from 20 years of binary software-as-a-medical-device testing to a framework assessing clinical performance. This shift addresses the challenge of validating non-deterministic generative outputs that lack the fixed parameters of legacy digital tools.

The players

FDA

The federal agency overseeing medical device safety, which has regulated software-based tools for two decades.

Rick Abramson

The director of the FDA's Digital Health Center of Excellence.

American Medical Association

A national organization representing physicians that provided feedback on frameworks for agentic and generative AI tools.

The details

The FDA is considering a regulatory approach that parallels the competency evaluation of human physicians rather than static code verification. This process focuses on defining safety boundaries within an unbounded evaluation space, where generative models produce varied, non-repeating results. By evaluating these systems as diagnostic or clinical partners, the agency aims to ensure that AI output aligns with professional clinical judgment.

October 19, 2026: Public comment period on GenAI discussion paper closed.
October 20, 2026: Rick Abramson discussed regulatory policy in Boston.
Upcoming months: Expected issuance of draft policy guidance.

The Tech Race

The agency's new focus on generative AI marks a significant shift from the binary testing protocols used throughout its 20-year history of regulating software as a medical device. This approach mirrors current efforts by international regulators to establish clinical safety standards for high-stakes autonomous systems.

Medical device developers will eventually face new compliance benchmarks centered on clinical judgment instead of simple code output validation. Once the draft guidance is issued, companies will have a clearer roadmap for the safety data required to bring generative AI tools to market.

The takeaway

The FDA is moving toward a physician-style competency model for software as it balances safety with the inherent non-determinism of generative tools. Stakeholders should monitor the coming months for the release of the draft policy guidance, which will define the new testing requirements.