Lauren Karkazis
Associate Editor
Loyola University Chicago School of Law, JD 2028
On August 18, 2026, the Food and Drug Administration (FDA) released a discussion paper and request for feedback on the regulation of generative AI-enabled medical devices. The paper discusses the regulatory considerations specific to generative AI-enabled medical devices, including monitoring after a device is deployed, evaluation of underlying data models, and multiple risk assessments. To support its regulatory efforts, the FDA proposes implementing two processes into their oversight model: a two-axis risk evaluation framework to assess the risk a continuously adapting device may pose and a competency-based approach to measure a device’s effectiveness. Although the discussion does not set forth any proposed regulations, it does invite early feedback from stakeholders. However, the FDA fails to adequately invite and consider feedback from doctors and patients interacting with generative AI-enabled medical devices.
The FDA proposes how to regulate generative AI-enabled medical devices
Generative AI-enabled medical devices’ ever-changing nature makes it challenging to regulate, especially after device approval. Combating this challenge is the reason for the paper’s focus on the FDA’s two key ideas for evaluating and regulating generative AI-enabled medical devices: a two-axis risk evaluation framework and a competency-based approach to measure device effectiveness. The two-axis risk evaluation framework weighs the likelihood of adverse outcomes alongside their potential severity. In doing so, the FDA hopes to adjust the scope of its regulatory oversight to cover constantly evolving devices, including its expectations for monitoring after a device enters the market. The competency-based approach would work similarly to assist in assessing doctors, evaluating a device’s analytical processes, and determining a device’s effectiveness in a clinical setting.
In order to effectively regulate generative AI-enabled medical devices under the proposed evaluation methods, the FDA needs data against which to measure the devices. This is called device benchmarking, the process by which a device demonstrates its clinical knowledge, analytic capabilities, communication, and behavior to support a reasonable assurance of safety and effectiveness in its intended use. The FDA does not yet describe its plan for sourcing this data; instead, it’s inviting industry advice on the best methods to use for data collection. The FDA also discusses additional possibilities for regulating generative AI-enabled medical devices, such as review by independent third parties, continued monitoring of devices after release on the market, and clinician involvement. For example, a panel of qualified physicians may evaluate how the device compares to the standard of care of traditional treatment.
The FDA needs to put more weight on patient and physician viewpoints
Throughout the discussion paper, the FDA proposes discussion questions and invites feedback from industry experts, device manufacturers, physicians, and other interested parties. It is critically important that the FDA consider feedback from a diverse range of viewpoints, especially patients, a group not mentioned in the paper’s requests for comments. Physicians and patients are uniquely situated to provide input on the two-axis risk framework and competency-based approach since they are the end users of generative AI-enabled medical devices.
Physicians are the leading experts and operators of medical devices. While the discussion paper does invite their involvement, it is critical that greater weight be placed on their opinions throughout the regulatory process. Physicians can provide valuable insight into the practical and ethical considerations that must be weighed when using these products, such as adjusting the standard of care in complex patient cases. Patients can provide user feedback into the practicalities of using these devices, such as ease of use for managing a medical condition. To ensure that generative AI-enabled medical devices serve the needs of patients and providers, the FDA must give their feedback real weight when developing future regulatory guidelines. To ensure that their voices are heard, the FDA should continue the comment period to a later date, survey patients and physicians who already use generative AI-enabled medical devices and maintain regular outreach to physician groups. While releasing the discussion paper is a step in the right direction, more targeted communications will be necessary to ensure that patient and physician interests are prioritized as the regulatory process advances; failing to take these valuable viewpoints into account will leave the two most impacted groups behind.