Medical School Curriculum Reform and the Missing Debate on Artificial Intelligence as an Educational Agent

In 2024, revision of the Enforcement Decree of the Higher Education Act in Korea abolished the rigid division between the 2-year premedical phase and the 4-year medical phase, allowing medical schools to design 6-year curricula autonomously for the first time in a century. Universities nationwide have begun redesigning their programs accordingly. This historic reform coincides with the rapid spread of generative artificial intelligence (AI) across society, including the ways students learn. In the current curriculum reform discourse, however, AI is discussed almost exclusively as content to be taught or as a supplementary tool to be used, whereas the more fundamental question of AI as an educational agent remains largely unexamined. Outside the formal curriculum, AI tutors already function as de facto teachers: a randomized experiment suggests that well-designed AI tutoring can outperform expert- led active-learning classrooms, and international surveys indicate that medical students routinely consult AI before, or instead of, their professors. This paper argues that the widening gap between the official, faculty-centered curriculum and the actual AI-mediated learning ecosystem may create discord: AI-mediated learning may become an unmanaged informal curriculum, faculty-designed assessment may lose authority, and trust in the formal curriculum may erode. Curriculum reform that merely rearranges the 6-year structure without addressing who teaches risks producing an elaborate design for an education that no longer reflects how learning occurs. Medical schools should explicitly define the position, scope, and governance of AI as an educational agent within their new curricula.