인공지능 관련 사고와 민사책임 - 불법행위책임을 중심으로 -

The Framework Act on the Development of Artificial Intelligence and the Establishment of a T rus t-Based E nvironment (hereinafter t he “AI Framework A ct”) e ntered i nto force on January 21, 2026. The AI Framework Act primarily targets high-impact artificial intelligence, generative artificial intelligence, and high-performance artificial intelligence for regulation, and in particular reflects concerns regarding the risks posed by high-impact and high-performance AI systems. Nevertheless, the AI Framework Act is closer in nature to a promotional statute than to a regulatory one, and as a regulatory framework it exhibits certain limitations. Regulation of risk may be achieved not only through ex ante or ex post administrative regulation, but also through tort law. Where administrative regulation of risk is incomplete or insufficient, the tort liability regime is expected to play a more active role. This article seeks to explore an appropriate tort-law response to accidents involving artificial intelligence. First, because high-impact and high-performance AI systems are likely to entail substantial risks, it is appropriate to impose strict liability (risk-based liability) on AI developers. At the same time, in order to ensure that the introduction of strict liability does not hinder innovation in AI technology, complementary mechanisms such as liability insurance and compensation funds should be established. Second, with respect to other categories of AI, it is difficult to conclude that they pose risks sufficiently serious to justify the imposition of strict liability, and the application of a fault-based liability regime should be adequate. In this context, further examination is required concerning the concept of fault, the allocation of liability among multiple responsible parties, and the mitigation of the burden of proof.