FDA Seeks Feedback on Potential Regulation of GenAI Medical Devices
The U.S. Food and Drug Administration (FDA) has issued a discussion paper on considerations for the regulation of Generative AI (GenAI)-enabled medical devices. As a regulator of all medical devices, the FDA is considering whether new regulations are required for GenAI-enabled medical devices to ensure patients are provided with timely access to safe and effective devices.
GenAI-enabled medical devices have the potential to transform patient care, yet the devices may introduce unique risks compared to traditional software and artificial intelligence (AI)-enabled medical devices. Current regulatory frameworks, such as those used for traditional medical devices, may not be appropriate for GenAI-enabled devices, which present unique challenges and risks.
The devices have unique characteristics and behaviours, including the capability to produce variable outputs, which change over time as the devices incorporate continuously learning systems. GenAI devices can accept open-ended inputs, and it is not feasible to test the full range of inputs and assess outputs using traditional premarket testing methodologies. The FDA notes that many devices are built on general-purpose foundation models, which have been developed by third parties that have varying levels of transparency into training data, architecture, and evaluation methods. As such, specific behaviors and errors are difficult to attribute to the underlying model used by the devices.
While GenAI-enabled devices offer a wide range of benefits over and above what can be gained from traditional and AI-enabled medical devices, the characteristics that provide those benefits also present unique risks. For instance, GenAI systems may misinterpret or distort data, filling in knowledge gaps with plausible but invented information (confabulations), such as associating a symptom with the wrong condition. There is also a risk of hallucinations – the generation of false facts – on which output is based and presented as fact. GenAI tools may provide outputs that are plausible and sound authentic to end users, which may be questionable at best and potentially dangerous to health.
The discussion paper, published by the FDA’s Center for Devices and Radiological Health (CDRH) – Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback – delves into the challenges associated with premarket evaluation and postmarket monitoring of GenAI-enabled medical devices. The aim of the paper is to guide discussion and focus feedback ahead of the potential development of guidance and future regulations. No decision has been made about whether FDA regulation is required, or areas where guidance may be necessary. The feedback obtained in response to the discussion paper will guide future FDA decisions, including new methodologies for premarket evaluation and the postmarket assessment of the performance of GenAI-enabled medical devices to ensure they remain safe and effective throughout the entire product lifecycle.
The paper discusses the possibility of competency-based testing of GenAI-enabled medical devices for premarket evaluations, using an approach modelled on medical training, licensure examinations, supervised practice, periodic reevaluation, and public reporting, and device benchmarking to assess whether a device demonstrates the necessary clinical knowledge, analytic capabilities, safety behavior, communication, and generalizability to support reasonable assurance of safety and effectiveness of the device for its intended use. Potentially, clinical confirmation will be required, as a competency-based approach may not fully assess performance in a clinical setting.
The FDA anticipates a risk-based approach will be necessary for regulation, taking into consideration the intended use and device characteristics. For instance, certain action-directing functions may be classed as higher risk than functions that provide non-directive information, as well as agentic AI systems capable of autonomous actions.
The FDA said it wishes to work collaboratively with the full range of stakeholders to develop efficient, scientifically sound, and least burdensome approaches to the premarket evaluation and postmarket monitoring of GenAI-enabled devices. Feedback on the discussion draft is requested from medical device manufacturers, clinicians, researchers, and the general public by October 19, 2026.
