Call for Licensing Medical GenAI Like Healthcare Professionals
A growing chorus of academic physicians, policy experts, and public health specialists is advocating for the licensing of medical Generative AI (GenAI) models in a manner akin to traditional healthcare professionals, such as doctors and nurses.
The Background
In November, Eric Bressman, MD, MSHP, a hospitalist at UPenn, along with colleagues from prestigious institutions including Harvard, Brown, and the University of Potsdam, published an opinion piece in JAMA Internal Medicine that put forth this notion. They suggested that amid the current uncertainty surrounding the AI regulatory environment, there exists an opportunity to develop a more agile and innovative framework for clinical AI.
“A licensure framework may help ensure that innovation scales with accountability and not ahead of it,” the researchers articulated, highlighting the need for a structured approach.
Recent Developments
Julia Hinkley, JD, director of policy strategy at UPenn’s Leonard Davis Institute of Health Economics (LDI), has added momentum to this discourse through a recent blog post published on January 29. She proposed that an ideal federal digital licensing board should oversee this new framework. Hinkley indicated that existing federal and state bodies could also play crucial roles in this regulation.
“The FDA could retain its role in premarket assessments, preventing developers from needing to submit to 50 state licensing authorities,” she suggested, advocating for a streamlined process.
Implementation Centers and Oversight
Hinkley also supports the idea of allowing health systems with AI expertise to function as “implementation centers,” which could facilitate the integration and regulation of AI technologies in healthcare.
Furthermore, state medical boards would provide ongoing oversight, collaborating with or deferring to a federal coordinating body to harmonize standards across the board.
The Need for Regulatory Innovation
Hinkley emphasizes that regulatory innovation is essential in this context. She draws parallels between the concerns surrounding generative AI—such as hallucinations and performance drift—and historical worries from the late 19th century regarding quack remedies and inconsistent clinician training.
“Licensure’s approach, combining practice standards with ongoing surveillance and education, can be adapted for AI regulation,” she notes, suggesting that a structured licensing framework could help mitigate potential risks associated with AI technologies.
Conclusion
Both the peer-reviewed paper and Hinkley’s blog post present a reader-friendly table that illustrates parallels between clinician licensing and a potential future licensing structure for AI. As the conversation around the regulation of medical GenAI continues to evolve, the call for a comprehensive licensing framework grows louder, advocating for accountability and enhanced patient care in the age of artificial intelligence.
