August 21, 2026
2 min read
Key takeaways:
- The FDA released a discussion paper detailing a conceptual framework for regulation of medical devices with generative AI features.
- Feedback on the framework is due by Oct. 19.
The FDA is seeking feedback on a potential way of regulating medical devices equipped with generative AI features, informed by the process of training human clinicians.
Described as a “competency-based approach,” the process would evaluate AI-equipped devices in increasingly clinical scenarios, including potentially with real patients.
The FDA is seeking feedback on a potential way of regulating medical devices equipped with generative AI features, informed by the process of training human clinicians.
It is “inspired, at a high level, by how human clinicians are evaluated and credentialed,” according to an FDA discussion paper.
The paper also discussed ongoing evaluation of AI devices after they come to market, noting that the FDA is “considering whether it is appropriate to accept greater premarket uncertainty … through greater reliance on postmarket monitoring.”
The paper said it is intended “for discussion purposes only” and is “meant to seek early input” from outside the FDA. In a statement to Healio, the agency said it intends to work collaboratively on “efficient, scientifically sound and least burdensome approaches” to evaluating generative AI-enabled medical devices “across the total product life cycle.”
“By inviting input from the public, we are launching a transparent process to inform the development of an approach that safeguards patients and consumers, advances innovation and serves as a potential model for regulators around the world,” the agency stated.
The deadline to submit feedback is Oct. 19.
In the paper, the FDA outlined a potential “competency-based approach” for premarket evaluation consisting of two parts: nonclinical benchmarking and clinical confirmation.
Benchmarking would assess the safety, clinical proficiency, generalizability and agentic AI capabilities of the devices — but not necessarily all four elements for each device. This could potentially be done using “synthetic data generation and simulation tools, including virtual patient avatars,” the FDA wrote.
Clinical confirmation would evaluate how AI-enabled devices perform “under real-world or clinically representative conditions.” Aside from prospective clinical studies, the FDA suggested this could be done through retrospective analysis of real patient data; “shadow deployment” in a live clinical workflow without affecting actual care; the use of patient actors; or clinician review of real patient-device encounters.
The FDA cited several research papers that informed these ideas. A coauthor of one of these, David Blumenthal, MD, MPP, professor at the Harvard T.H. Chan School of Public Health, told Healio the FDA had offered “a thoughtful and useful beginning of a conversation.”
David Blumenthal
“It’s not settled policy, and the framework could evolve considerably during the process of collecting information,” he said. “There’s a lot of work that needs to be done on the clinical confirmation side and that needs to be piloted and worked out in collaboration between healthcare professionals and AI developers.”
While commending the FDA for having “an open mind and seeking broad input,” he added it may be “operating in areas where they don’t yet have authority.”
“It’s not clear whether FDA will actually be an agency that implements the process they have outlined because it’s not always clear that generative AI used in healthcare is actually a medical device,” he said. “What they understandably didn’t address, because they are part of a government that has complicated processes, is the need for new legislation, if any; for new agencies, if any; for private sector capabilities, if any.”
For more information:
David Blumenthal, MD, MPP, can be reached at dablumenthal@hsph.harvard.edu.
Sources/Disclosures
Source:
References:
Disclosures:
Blumenthal reports being on the board of directors for Aledade.
Ask a clinical question and tap into Healio AI’s knowledge base.
- PubMed, enrolling/recruiting trials, guidelines
- Clinical Guidance, Healio CME, FDA news
- Healio’s exclusive daily news coverage of clinical data
