FDA took in more than 500 comments on its August discussion paper on generative AI enabled devices, filed under docket FDA-2026-N-7874. The comment period closed in mid-October. The paper floated evaluating these devices with a framework modeled on physician credentialing, and leaned heavily on postmarket monitoring.
Industry asked the agency to extend what already exists. AdvaMed said adding generative AI should not by itself trigger a competency assessment, and called the predetermined change control plan mechanism "too restrictive" for generative models. It also argued postmarket monitoring should be risk-based and carried by manufacturers. The Medical Device Manufacturers Association and GE HealthCare took the same position.
Clinicians and patient groups pushed the other way. Interventional cardiologists writing for SCAI said FDA has to separate decision support from decision making, a line generative AI blurs. Generation Patient backed wider postmarket monitoring but said it cannot stand in for premarket evidence. Stanford's Brainstorm lab asked FDA to test performance against reasonably foreseeable use rather than the labeled indication alone, and to judge high-consequence failures such as missed sepsis or wrong insulin dosing on their own pass or fail standard.
Oura asked FDA not to treat informational, non-directive outputs as devices at all. Hospitals and provider groups argued that developers, not treating physicians, should answer for how these devices perform.
If the agency takes the change control route industry wants, the work lands in your PCCP. Write it now on the assumption that model updates get reviewed against foreseeable misuse, not just your indications for use statement.