Review of the 2025 Fall Conference of the Korean Society of Medical Informatics: Generative AI in Healthcare Systems—From Insight to Impact

The rapid advancement of generative artificial intelligence (AI) is reshaping healthcare systems worldwide, signaling a shift from exploratory innovation toward tangible, systemlevel implementation.Within this evolving landscape, the 2025 Fall Conference of the Korean Society of Medical Informatics (KOSMI), held from November 22 to 24 in Incheon, provided a timely and substantive forum for examining how generative AI can be translated into sustainable impact across healthcare systems.Under the theme "Generative AI in Healthcare Systems: From Insight to Impact, " the conference convened 1,131 participants from clinical, academic, industrial, and policy sectors, underscoring the increasingly interdisciplinary character of medical informatics.The conference opened with remarks by Woo-Kyung Kim, Chair of the Organizing Committee and President of Gil Medical Center, followed by a welcome address from Byung Chul Chang, President of KOSMI.Both speakers emphasized the responsibility of the medical informatics community to guide the integration of generative AI beyond tech-nological novelty and toward system-level transformation.In congratulatory remarks, Minseob Yeom, President of the Korea Health Information Service, highlighted the central importance of national health data infrastructure, standardization, and governance in enabling trustworthy and scalable AI adoption.A central highlight of the opening program was the Beom-San Special Lecture delivered by Sang-Hoon Jeon, Professor at Seoul National University Bundang Hospital.His lecture emphasized that clinical insight, domain expertise, and informatics infrastructure must co-evolve for generative AI to function as a reliable component of real-world healthcare systems.Complementing this perspective, the keynote lecture by Ming-Chin Lin, Professor at Taipei Medical University, underscored the importance of human-centered AI and cross-disciplinary collaboration, emphasizing that the primary challenge of medical AI lies not in technical feasibility but in responsible integration that preserves human oversight, accountability, and trust.The scientific program was comprehensive and carefully structured, comprising four tutorials, 23 symposia, 10 oral presentation sessions with 48 presentations, 87 poster and e-poster presentations, and three industry-supported symposia.Across these sessions, a gradual shift in emphasis was evident.Rather than focusing exclusively on isolated algo-