Hospital chief executives must take personal responsibility for AI safety and performance, rather than delegating oversight to IT departments or subcommittees, three healthcare leaders argue in a STAT opinion piece published July 22.
Peter Pronovost of University Hospitals, Justin Norden of Qualified Health, and Kedar Mate warn that most health systems are monitoring AI tools the same way they governed MRI scanners in 2010: a subcommittee, a checklist, and a quarterly meeting. That approach can take six months or longer and is dangerously inadequate for AI systems that continuously evolve and affect clinical decisions in real time.
AI is already drafting clinical notes, flagging sepsis, screening imaging, processing prior authorizations, and answering patient messages in most US hospitals. But few organizations have the continuous monitoring infrastructure needed to detect when model performance drifts or when new data changes an algorithm’s behavior. The authors argue that AI governance must become a board-level priority with clear accountability structures.
The piece adds to growing calls for stronger AI oversight in healthcare. Recent surveys show most health systems lack formal AI governance processes, and federal agencies including the FDA and CMS are beginning to establish clearer expectations for how AI tools should be evaluated and monitored in clinical settings.
