자율형 의료 AI에 대한 제조물책임과 책임배분 – Watson 이후 의료 AI와 비교법적 시사점 -

초록

With the rapid advancement of science and technology, the risks associated with highly autonomous medical AI are increasingly present in clinical settings. Autonomous medical AI does not merely assist healthcare professionals in their decision-making; it directly impacts patients in actual clinical settings. Consequently, the focus of risk shifts from the negligence of individual healthcare professionals to the systemic level—a combination of AI algorithm design, training data, warning mechanisms, and update systems. Previous studies have generally addressed these risks as a question of who should bear the responsibility. However, the unique nature of autonomous medical AI lies in the “timing” of these risks. The Product Liability Act assesses defects based on the time the product was supplied, and defenses such as the “development risk defence” defense are also premised on that same point in time. However, with autonomous medical AI, the risk structure changes through updates made after deployment. This paper proposes an “integrated model of liability for autonomous systems” applicable to South Korea. It also reconceptualizes the role of healthcare professionals in a manner distinct from previous approaches and seeks to ensure effective redress for victims by considering the introduction of liability insurance for medical institutions and manufacturers, as well as a no-fault advance compensation system. Thus, moving beyond the debate over whether software should be included as a “product” in the context of product liability, this paper reconceptualizes the timing of defect determination and proposes a new framework for the burden of proof and allocation of liability to ensure that product liability functions effectively.

키워드