Should AI Doctors Be Licensed? STAT Pushes a Framework for Autonomous Clinical AI
A STAT opinion argues that autonomous clinical AI should be licensed, proposing a formal framework for systems that move beyond decision support. The idea reflects a growing recognition that the current patchwork of oversight may be inadequate for high-stakes AI used in patient care.
The licensing idea is provocative because it frames autonomous clinical AI less like software and more like a profession with privileges and responsibilities. That may sound extreme, but it reflects a real regulatory gap: tools increasingly influence diagnosis, triage, and treatment planning without fitting neatly into existing medical-device categories.
A licensing framework could help clarify expectations around competence, auditing, monitoring, and accountability. In practice, that might mean evaluating not only model performance, but also the conditions under which a system can be deployed, the types of clinicians who must supervise it, and the post-market surveillance required to keep it in use.
The challenge is that AI is not a static practitioner. Models change, data drift, workflows evolve, and performance can vary dramatically across settings. A license for a one-time product approval may be ill-suited to a system that must be continuously monitored and recalibrated.
Still, the broader point is hard to ignore: if AI systems are going to act with increasing independence in healthcare, the oversight model will need to evolve accordingly. Whether that becomes formal licensing or some other regime, the status quo of one-and-done approval looks increasingly out of step with clinical reality.