의료 인공지능과 민사책임에 관한 연구-의료과실 책임과 제조물 책임을 중심으로-

Medical artificial intelligence (AI) based on machine learning and deep learning has driven substantial innovation into healthcare, improving accuracy, speed, and efficiency. However, it also carries the risk of error stemming from data bias and algorithmic imperfections. When a patient suffers harm due to an incorrect AI prediction, liability typically concerns both the physician who relied on the prediction and the AI system’s manufacturer. Currently, medical AI functions as a decision-support tool, imposing on physicians the duty to assess the appropriateness of AI recommendations in light of prevailing medical standards. A breach of this duty may constitute medical malpractice. In the case of nearly-fully autonomous AI, the physician-AI relationship may shift toward a horizontal division of medical labor, allowing for the potential application of the principle of trust among actors. In product liability claims, AI’s complexity, opacity, unpredictability, and autonomy make it difficult for plaintiffs to prove the feasibility of reasonable alternative designs, thereby lowering the likelihood of success. The revised EU Product Liability Directive includes software within the definition of “product” and introduces measures such as evidence disclosure obligations and presumptions of defect to ease the plaintiff's burden of proof. Nonetheless, substantial limitations remain, particularly regarding autonomous AI systems. This study argues that assigning sole liability to physicians is undesirable for both patient protection and technological innovation. It explores alternative frameworks, including AI legal personhood, common enterprise liability, and strict (no-fault) liability. Drawing on the liability insurance model for autonomous vehicles, it proposes mandatory no-fault liability insurance for physicians or hospitals employing medical AI.