FDA Signals Upcoming Generative AI Device Guidance: What the Discussion Paper Means for Companies

The U.S. Food and Drug Administration has published a discussion paper outlining how its existing medical device framework may apply to products using generative AI. The paper does not create new requirements but represents a formal step toward guidance, with the FDA's director of its Digital Health Center of Excellence confirming that "formal policy guidance" is the agency's goal.

The FDA has already been active in this space. In 2026, it granted Breakthrough Device Designation to Cognita Chest X-Ray (a generative vision-language model for radiology report drafting) and Aidoc First Read (a chest X-ray analysis tool). Both products generate substantive clinical text for physician review.

Key takeaways:

• FDA is not creating a new framework but is exploring how to apply existing risk principles to GenAI's variable outputs
• A proposed two-axis framework considers: (1) the form and independence of the device's activity; (2) the potential consequences of an incorrect output
• FDA is exploring a competency-based approach to premarket evaluation covering clinical knowledge, reasoning, refusal behavior, and reliability
• Post-market monitoring may be required because some GenAI devices are difficult to fully evaluate before marketing
• Third-party foundation models create new change-control challenges: manufacturers may not control when or how underlying models change
• FDA is questioning whether patient-facing GenAI functions present greater risk than provider-facing functions due to automation bias
• Stakeholders can submit comments under docket FDA-2026-N-7874 by October 19, 2026

Companies developing GenAI devices should review their intended use definitions, validation strategies, and third-party model contracts now, ahead of formal guidance.

Read the full article on Mondaq / Buchanan Ingersoll and Rooney PC