Wow…this may be the biggest drop in regulatory arena this year, as the FDA Seeks Public Feedback to Inform Regulatory Approach for Generative AI-Enabled Medical Devices.
I strongly encourage anyone in the FDA regulated device space currently, considering entering the FDA regulated space, or an interested healthcare professional (HCP) or the general public to take a close look at the discussion paper and submit feedback by October 19, 2026 at the public docket. The stakes of this FDA request couldn’t be higher and will clearly shape how Gen AI enables devices will be regulated by the FDA.
First off, I want to congratulate the FDA for clearly taking stock of their prior 2024 and 2025 Digital Health Advisory Committee (DHAC) meetings to inform this recent discussion paper. This discussion paper dropped today is proposing some bold elements for regulatory consideration.
I’ll high level comment on the four major portions of this document that people need to consider. I encourage folks to take a close look, as there is a lot of detail to this discussion paper, and comment at public docket when possible.
Considerations for the Assessment of Risk for GenAI-Enabled Devices — the FDA is considering framework for assessing risk across a 2 axis (activity vs. consequence) starting at informational: non-directive (lowest risk) and continuing upwards on risk in informational: action-directing, action-taking: HCP-supervised and last action taking: fully autonomous (highest risk).
A Competency-Based Approach for Premarket Evaluation of GenAI-Enabled Devices — the FDA is considering a draft competency-based approach with device benchmarking anchors to evaluate performance of devices. Clinical confirmation of performance depending on the device could take the form of the following progression through increasing rigor and patient exposure: retrospective evaluation on real patient inputs (least rigor / patient exposure), shadow deployment, standardized patient interactions, clinician adjudication of real cases and last prospective clinical study (most rigor / patient exposure). Standards of performance are a key consideration. The potential roles for independent third parties, either in Accreditation Scheme for Conformity Assessment (ASCA) accredited testing labs, or using the FDA Medical Device Development Tool (MDDT) program were also considered.
Postmarket Monitoring for GenAI-Enabled Devices — the FDA is considering whether it is appropriate to accept greater premarket uncertainty regarding a GenAI-enabled device’s benefit-risk profile through greater reliance on postmarket monitoring. This in itself is a remarkable proposal, but a practical one, and the FDA is considering postmarket monitoring in periodic device benchmarking, periodic sample-based clinician review and performance degradation monitoring. The FDA recognizes that this postmarketing monitoring would be a shared ecosystem responsibility that all participants may play some role. Predetermined change control plans (PCCPs) could be part of the postmarket monitoring picture if unexpected outcomes / drift / degradation are observed, to facilitate certain device changes without requiring a new premarket submission.
Foundation Model Device Master Files (MAF) and Agentic AI Considerations — FDA is considering the feasibility of voluntary Foundation Model MAFs, leveraging the existing Device Master File program, under which foundation model developers and platform providers could voluntarily submit information such as structured model cards or system cards to FDA. MAFs would be held by FDA confidentially and could be referenced by sponsors, with the file holder’s authorization, in support of individual premarket submissions. Agentic architectures may raise additional considerations beyond those applicable to other GenAI-enabled devices that may warrant different approaches to evaluation.
I am hopeful that this comment period will result in a future DHAC meeting that would be open to the public to attend and comment. The two prior DHAC meetings chaired by Dr. Ami Bhatt were clearly foundational to getting us where we are now, and for that I am thankful. But if another public DHAC meeting doesn’t convene, comments made to this docket are even more critical, as the next step by FDA could be the issuance of draft guidance(s) based off of commentary.
