The U.S. Food and Drug Administration is asking manufacturers, clinicians, researchers, and the public to help shape how generative AI-enabled medical devices should be regulated. The agency released a discussion paper on Aug. 18 covering risk assessment, premarket review and postmarket monitoring, with comments due Oct. 19, 2026, under docket FDA-2026-N-7874.
The FDA is considering a two-axis approach to risk assessment that could help determine the level of regulatory scrutiny required for different GenAI medical devices. For premarket review, the paper proposes exploring a competency-assessment model that would combine non-clinical benchmarking with clinical confirmation to establish whether a device performs as intended before it reaches patients.
The agency is also examining how these systems should be monitored after authorization. Its discussion includes risk-proportionate postmarket oversight and regulatory questions surrounding foundation models and agentic AI, reflecting concerns that newer AI systems can behave differently from conventional medical software.
The proposal comes as AI-enabled medical devices are already moving through the FDA’s existing authorization system. The agency maintains a public list of AI-enabled devices that have received authorization for marketing and met applicable premarket requirements, including reviews related to safety, effectiveness, intended use and supporting studies.
Recent entries show the breadth of that market. The FDA list includes devices such as AiORTA – Plan v2.0, finalized March 30, 2026; ECG-AI Pulmonary Hypertension 12-Lead algorithm, finalized March 28; and several AI-enabled radiology systems finalized March 27, including Spectral CT Verida Family, Butterfly Gestational Age Tool and Automated Aortic Stenosis Software.
But the existing list has an important limitation: the FDA says it is not comprehensive. Devices are identified primarily through AI-related terminology in authorization documents or device classifications. The agency is therefore exploring ways to identify and tag devices using foundation models, large language models and multimodal architectures, which could give clinicians and patients clearer visibility into how these technologies are being used.
That distinction matters because the new discussion paper does not establish new GenAI regulations. The FDA explicitly says the paper is for discussion purposes and is neither draft nor final guidance. It does not establish new policy, set final regulatory expectations or determine what legal authority would be required for any future framework.
The effort is being led by the FDA’s Digital Health Center of Excellence within the Center for Devices and Radiological Health. The agency says the goal is to develop a regulatory approach that can address the specific risks of generative AI while supporting the development of safe and effective digital health technologies.
The FDA is accepting feedback from device manufacturers, clinicians, researchers, consumers and other interested parties through Oct. 19. Their responses will help inform the agency’s next steps as generative AI moves further into medical-device development and clinical care.
Bottom Line
The FDA has opened the regulatory conversation, not finalized new GenAI rules. Its focus is on deciding how risk, premarket evidence, and ongoing monitoring should work for medical devices built with increasingly capable generative AI, while existing AI-enabled devices continue to enter the U.S. market through established FDA pathways.
