의료 인공지능 활용에 관한 책임 논의 : 미국과 유럽연합의 규제 현황을 중심으로

With the development of AI technology, medical AI technology is being used in various areas of the medical field, and the scope of use is gradually expanding. However, due to the characteristics of the black box of AI, there is a problem of a responsibility gap that is difficult to clarify responsibility with the existing legal system. Moreover, since medical AI is used as a tool directly related to human life and health, the discussion of responsibility should be clearly organized before use. Therefore, in this paper, in order to propose implications for the problem of distributing responsibilities according to the use of medical artificial intelligence, improvement plans and implications were derived by analyzing the regulatory status of the US and the EU, which are major countries. First of all, before discussing the legal responsibility of AI, we looked at whether AI can be recognized as a legal entity. Although AI autonomy is strengthened with the development of AI technology, artificial intelligence autonomy is only functional and behavioral autonomy and is different from human moral and legal autonomy. Therefore, medical AI cannot be the final decision maker as an auxiliary tool that assists and strengthens doctors' judgment, and medical AI should be used as a human-centered approach. Utah and the EU AI Act are taking a joint responsibility approach by stipulating that manufacturers, medical institutions, and users/nature persons share responsibilities. The United States is focusing on the use and promotion of artificial intelligence, so it is taking a method of defining medical institutions as the primary responsibility and giving doctors the final responsibility for decisions. On the other hand, the EU, which focuses on safe use, stipulates manufacturers as a way to strengthen their life cycle responsibility obligations. Although there may be pros and cons between the two systems, considering the usefulness of medical artificial intelligence technology, laws and systems should be prepared to achieve a balance between utilization and safety, considering that excessive regulation hinders technological development. Accordingly, clear new norms on responsibility for the use of medical AI need to be established, and it is judged that the regulatory method is appropriate to hollow out responsibility and establish a distributed regulatory system. In addition, software and algorithm-based medical AI devices need to be viewed as product targets, as in the EU's product liability guidelines, and it is necessary to clarify the manufacturer's responsibility by supplementing the liability attribution structure and proof standards. However, if this also focuses on the use of medical AI, elaborate legislation should be implemented so as not to become a wall of excessive regulation.