The FDA released a discussion paper on August 18 outlining how it might eventually regulate generative AI-enabled medical devices, opening a public comment docket that runs through October 19. The paper comes from the agency’s Digital Health Center of Excellence, now led by Rick Abramson, M.D., who took over the role in February.
The FDA has been explicit that this is a discussion paper, not draft or final guidance, and doesn’t represent a policy change. It’s meant to gather input from outside groups before the agency commits to an actual regulatory framework, covering questions like risk assessment, premarket evaluation, and how to monitor an AI model’s performance after it’s already in clinical use.
The Core Problem the FDA Is Trying to Solve
Traditional medical device regulation assumes a product’s behavior is fixed once it’s approved, but generative AI models can update, retrain, or drift in how they respond over time, sometimes in ways their own developers can’t fully predict or explain. The discussion paper proposes a two-axis risk framework and a “competency assessment” approach, pairing non-clinical benchmarking with clinical confirmation, as one possible way to evaluate a moving target rather than a static product.
What Happens Next
Public feedback collected through October 19 will inform whatever the FDA eventually proposes as actual draft guidance, a process that has no set timeline yet. In the meantime, hospitals and AI medical device makers are operating under existing device regulations that weren’t written with generative AI in mind, meaning the current gap between the technology’s pace and the agency’s rulemaking process is likely to persist for a while longer regardless of how this discussion period concludes.
