On August 18, 2026, the FDA issued a discussion paper on medical devices using generative AI; comments may be submitted until October 19, 2026, under docket number FDA-2026-N-7874. Nothing in it is binding; it is not guidance, but rather a collection of considerations and open questions.
The agency is proposing a two-axis risk heuristic for discussion:
how far-reaching and autonomous a product’s actions are—ranging from purely informational output to autonomous behavior—and how severe the damage would be if one were to rely on incorrect output.
Of interest for patient-facing applications: The FDA tends to view the line between providing information and guiding action as a continuum, and it is considering whether a phrase such as “talk to your doctor” actually makes a message less action-oriented. The line of reasoning is interesting: Anyone seeking to mitigate the impact of a message with such a disclaimer faces the same discussion when classifying it under the applicable MDR Rule 11.
With regard to validation, the paper outlines tests against predefined thresholds and clinical confirmation, which does not necessarily require a prospective study in every case. After market authorization, monitoring for performance degradation should then take place.
The FDA’s considerations can be found here: https://www.fda.gov/medical-devices/digital-health-center-excellence/considerations-regulation-generative-ai-enabled-medical-devices-discussion-paper-and-request