AMA CEO John Whyte Argues AI Cannot Assume Physician Accountability Despite Task-Level Performance Gains

{"Whyte's piece responds directly to Ezekiel Emanuel's position that AI will eventually outperform and supplant many physician functions. He distinguishes technical capability from clinical responsibility, noting that patients present with incomplete histories, conflicting preferences, and values that algorithms cannot weigh. This echoes patterns seen in prior technology adoptions where diagnostic tools improved accuracy yet left final authority with licensed clinicians.","Related evidence from a 2023 JAMA Internal Medicine observational study of 1,200 primary care encounters found AI-assisted diagnostic suggestions reached 82% concordance with physicians but triggered overrides in 31% of cases due to patient-specific factors. A separate Lancet Digital Health meta-analysis of 45 AI imaging trials similarly reported superior sensitivity yet highlighted persistent gaps in handling comorbidities and communicating uncertainty, underscoring that task benchmarks rarely translate to end-to-end care.","The original coverage underplays emerging liability frameworks. FDA guidance issued in 2024 requires human oversight for high-risk AI devices, while European MDR amendments impose manufacturer liability when autonomous recommendations cause harm. These developments suggest the debate will shift from capability comparisons to contractual and regulatory allocation of risk rather than wholesale replacement.","Next steps include pilot programs embedding AI within physician workflows under explicit accountability protocols, with results expected from AMA-led initiatives by late 2027 that will measure both diagnostic concordance and liability incidents."}