인공지능(AI)을 이용한 의료행위의 법적 쟁점 연구 - 의료행위의 주체성과 책임 구조를 중심으로 -

Artificial intelligence (AI) in healthcare has increasingly penetrated core stages of medical decision-making, including diagnosis, treatment selection, and prognosis prediction. As the accuracy and scope of medical AI systems continue to expand, fundamental legal questions arise regarding the subject of medical practice and the allocation of legal responsibility. In particular, the debate over whether medical AI can be regarded as an independent subject of medical practice poses significant challenges to the traditional structure of medical law. This article moves beyond the binary question of whether medical AI should be recognized as a legal subject of medical practice. Instead, it focuses on how medical AI can be institutionally integrated into the existing framework of medical liability while preserving the core principles of medical law. To this end, the article first examines the legal concept of medical practice under Korean law, the license-based structure of medical responsibility, and the civil and criminal liability of physicians. It then analyzes how the introduction of medical AI affects these established legal doctrines. Subsequently, the article conducts a comparative legal analysis of regulatory and liability frameworks for medical AI in the United States, the European Union, and Japan. This comparative study demonstrates that, despite differences in regulatory approaches, these jurisdictions commonly position medical AI as a high-risk medical technology that assists, rather than replaces, human medical judgment, while consistently maintaining that the final decision-making authority and legal responsibility remain with human physicians. Based on this comparative analysis, the article proposes a normative framework for reconstructing physicians’ duties of care and duties of explanation in the context of medical AI. It also examines the organizational responsibility of medical institutions and the relationship between physicians’ liability and product liability for medical AI manufacturers. Furthermore, the article suggests a phased legislative approach that reflects the dynamic and continuously evolving nature of medical AI technologies. In conclusion, this article argues that the central legal challenge in the era of medical AI does not lie in transferring the subject of medical practice from humans to technology. Rather, it lies in maintaining human physicians as the legal subjects of medical practice while refining liability structures to effectively absorb the new risks introduced by medical AI. This approach enables the legal system to accommodate technological innovation without undermining the public nature of medical practice or the protection of patients’ rights.