인공지능 규제 입법에 관한 일고찰

Abstract

In February 2023, the “Bill for the Act on Promotion of the Artificial Intelligence Industry and the Establishment of Trust” was approved by the Bill Review Subcommittee of the National Assembly’s Standing Committee. However, the Bill did not progress further and was automatically discarded as the conclusion of the 21st National Assembly in late May 2024. It was due to social criticism that the bill was insufficient to regulate AI risks. As social concerns about AI’s potential risks - such as safety threat, bias, discrimination, privacy infringements, opacity, and post-truth - grow, there is an increasing movement towards AI regulation, particularly in advanced regions such as the EU and the United States. Initially, AI regulation focused on securing AI ethics through soft law at the level of self-regulation. However, there has been a gradual shift towards imposing statutory regulation on AI-related operators. AI statutory regulation can be categorized into three main approaches. The first is a direct regulatory approach, where general legislation on AI governs common regulatory provisions to ensure the safety, ethics, and trust of AI technology as a whole. The second is an indirect regulatory approach, where legislation in other fields indirectly regulates the development and utilization of AI technology. The third is an individual regulatory approach, where legislation specific to certain sectors regulates the use of AI within those areas. As soon as the 22nd National Assembly opens, several bills to directly regulate AI were proposed. When introducing statutory regulation of AI, it is crucial to implement the minimal necessary regulation to mitigate specific risks in individual fields, while considering the potential growth and sustainability of AI technology. It is also essential to take into account international regulatory trends, as well as the domestic industrial ecosystem and governance realities. Therefore, from the perspective of direct regulation, it is desirable for general legislation of AI to promote self-regulation. This approach will further contribute to fostering technological advancement and industrial growth. However, to ensure a minimum level of transparency, it is necessary to impose statutory regulations that mandate the disclosure of various information by high-risk AI operators. In fields where the use of AI poses significant risks, it is advisable to adopt an individual regulatory approach with specialized and detailed provisions within the respective legislation.