JMIR Publications

ABSTRACT

Since the public release of large language models, generative artificial intelligence (GenAI) has moved rapidly from novelty to daily use in clinical medicine, and medical education is now under pressure to prepare physicians for a career that will be shaped in unpredictable ways by these new technologies. In 2025, the Carl J. Shapiro Institute for Education and Research at Harvard Medical School and Beth Israel Deaconess Medical Center, in partnership with the Macy Foundation and the Association of American Medical Colleges, convened teams from eight US medical schools at Millennium Conference 2025 to develop guidance for integrating GenAI into undergraduate and graduate medical education. In this Viewpoint, the conference writing group argues that the prime question facing the field of medical education is not primarily technical but rather a deeper philosophical question about how to cultivate the core responsibilities of physicians in a world shaped by powerful “thinking” technologies. Drawing on our conference deliberations, we discuss our genuine disagreements and points of debate, especially whether GenAI is a tool or a member of the clinical team, and arrive at the concept of the “humanistic scientist,” a physician whose judgment, communication, and responsibility cannot be delegated to an algorithm. These discussions involve creating spaces to protect AI-free practice, how to assess learners who always have AI at hand, and how to prepare faculty to teach with AI. We distinguish recommendations relevant to today from commitments that will endure for the foreseeable future. Finally, we contend that accountability, not capability, should govern how AI enters medical training.