Licensing clinical AI like it’s a doctor: 3 medical leaders issue 6-step call

Licensing today’s medical AI to “practice” alongside actual physicians would be an imperfect but, in many cases, perfectly appropriate way to oversee the technology without crimping its capabilities. 

What’s more, it would represent a solid upgrade over current FDA safety guardrails around AI-equipped medical devices. 

So believe three leading researchers who are close watchers of technology and medical ethics.

They make their case in an opinion piece published by the Journal of the American Medical Association.

“Rapidly advancing autonomous clinical AI presents a regulatory challenge that the FDA’s current framework, built for drugs and devices, poorly addresses,” the trio write. “That framework assumes static products, narrow indications, and clear manufacturer accountability, assumptions that autonomous clinical AI defies.”

Licensure, they add, is “better aligned with the realities of autonomous clinical AI.”

The authors are Alon Bergman, PhD, Robert Wachter, MD, and Ezekiel “Zeke” Emanuel, MD, PhD.

The scholars acknowledge that licensing AI would not settle questions around overarching concerns such as data governance and workforce transformation. These and other matters, they suggest, probably call for new regulation or legislation. 

However, they emphasize that, as clinical AI “increasingly resembles clinicians in its capabilities, our regulatory frameworks must evolve accordingly.”

Here are six key details in their proposal.  

 

1. Competency certification could begin with standardized examinations. 

Bergman and colleagues envision a future in which every autonomous AI clinical model is tested on all three components of the U.S. medical licensing examination (scientific principles, clinical knowledge and preparedness for decision-making).

Before being greenlit by the FDA, these models would have to beat or equal the median scores of recent human exam-passers. 

Each would also have to prove accurate and reliable on any relevant specialty board examinations aligned with the model’s intended scope. 

“Passing scores would establish minimum competency, not clinical readiness,” the authors write. 

2. Models that meet this threshold would enter a supervised clinical deployment period.

Such a period would emulate residency training, as models would have to demonstrate non-inferior clinical performance. 

“Specific requirements for patient volume would be established through the multistakeholder standards process,” the authors explain. 

3. This licensure-like approach would define a scope of practice.

In so doing, the process would specify which clinical functions specific AIs may perform, in which settings—and under what level of oversight. 

“An AI certified for primary care triage,” the authors offer as an example, “could gather histories and recommend next steps but could not prescribe care or medications without clinician approval.”

4. Certification would be time-limited.

It would also be contingent on ongoing performance monitoring and reevaluation, Bergman and colleagues propose.  

“Because AI systems are adaptive rather than static, continued authorization would depend on periodic (eg, biennial) demonstration of acceptable levels of clinical performance.”

5. Accountability structures would be explicit. 

AI developers would bear primary responsibility for model performance, the authors maintain. Performance measures would mandatorily include safety, accuracy and behavior across clinical contexts.

Meanwhile, deploying institutions would be responsible for implementation, the team suggests. Aspects of implementation would include integrating the AI into clinical workflows, monitoring patient outcomes and reporting adverse events, and ensuring that the conditions covered under the AI approval are followed—including triaging to a clinician when appropriate. 

“This layered structure mirrors existing liability frameworks in medicine,” the authors point out. 

6. Federal preemption would be necessary to avoid 50 parallel licensing regimens. 

“Under this approach, federal certification of AI competency would be binding on all states, whereas states would retain authority over scope of practice, supervision requirements and enforcement,” Bergman, Wachter and Emanuel write. 

More: 

‘Recognizing autonomous AI’s rapid evolution, this licensure-like oversight would subject autonomous clinical AI models to more continuous, comprehensive, flexible and clinical evaluation than the FDA’s current Software as a Medical Device framework.’

Hear them out in full here.