The Food and Drug Administration is considering evaluating generative AI-enabled medical devices through a competency-based approach similar to how doctors are evaluated, according to a discussion paper first shared with Axios. The agency says the paper is still in a preliminary phase, but the direction is clear enough: the federal apparatus wants a new way to sort which AI medical devices get through and which ones don’t.
Who Gets to Set the Standard
Michelle Tarver, director of the FDA Center for Devices and Radiological Health, said, "We are launching a transparent process to inform the development of an approach that safeguards patients and consumers, advances innovation and serves as a potential model for regulators around the world." That’s the language of control dressed up as openness. The agency is not stepping back from power; it’s redesigning how that power gets exercised over devices, patients, and the doctors who may or may not be in the loop.
The paper says the agency's "ultimate goal is to enable a nimble regulatory approach that employs least burdensome principles and allows patients timely access to safe and effective medical devices." That phrase, "least burdensome," does a lot of work. It signals a system trying to keep the gate while pretending the gate is lighter to push through. Patients are still the ones waiting for access, while regulators decide what counts as safe enough.
The Machine Meets the Bureaucracy
Acting FDA commissioner Kyle Diamantas said, "Artificial intelligence is transforming medicine, and the United States must lead in shaping how this technology is developed and used safely and responsibly." The paper frames the issue as a national leadership project, but the real story is the same old hierarchy: a federal agency claiming the right to define the terms of a technology that will reach ordinary people through hospitals, clinics, and devices they don’t control.
The paper says emerging medical technologies are challenging the existing framework for overseeing medical devices, and that the regulatory structure so far hasn't kept up. It says new technologies range from information tools to devices designed to provide patient care, either with or without a doctor in the loop. That gap matters because the people most exposed to the consequences are not the ones writing the framework. They’re the patients on the receiving end of whatever the system decides is acceptable.
Regulatory experts and the FDA note that generative AI devices are substantially different from other regulated medical devices because of variation in their outputs and because they evolve over time. That’s the problem the bureaucracy is trying to box in: software that shifts, changes, and refuses to sit still inside the old paperwork machine.
What the Agency Wants to Measure
The paper outlines a method for evaluating the risk of a GenAI device by measuring the type of activity performed by a device against the severity of the consequences from incorrect outputs. It discusses a competency-based approach inspired at a high level by how human clinicians are evaluated and credentialed, with benchmarks and performance confirmation in clinical settings. In other words, the FDA is looking at AI devices the way institutions look at workers they want to certify, monitor, and discipline.
It also weighs approaches to tracking how devices perform once on the market and says the agency 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. That’s the familiar bargain from above: let it in first, watch it later, and hope the damage stays within the lines.
The paper says that for many devices, performance might be compared to that of a panel of qualified clinicians whose consensus reflects the applicable standard of care, or to that of a median clinician in practice. So the benchmark becomes another institutional average, another managed standard, another way to turn messy human care into a metric the state can file and defend.
The agency says the paper raises considerations for both preapproval and postmarket regulation of AI-enabled devices. For now, it remains a discussion paper. But the shape of the system is already there: centralized authority, technical language, and a promise that the people affected will be protected by the same institutions deciding the rules.