인공지능기본법상 규제 샌드박스 제도화 연구 - EU 인공지능법의 규제 샌드박스와 디지털 옴니버스를 중심으로

This study analyzes the regulatory sandbox framework under the European Union Artificial Intelligence Act (EU AI Act) to identify institutional conditions for the effective implementation of risk-based AI regulation, and on this basis examines how an AI-specific regulatory sandbox could be institutionalized under Korea’s Framework Act on Artificial Intelligence. By reviewing regulatory developments following the EU AI Act and the Digital Omnibus package, this study shows that the EU regulatory sandbox is designed not merely as a temporary regulatory exemption, but as an institutional infrastructure that enables ex ante compliance assessment, controlled real-world testing, and regulatory learning for supervisory and legislative standard-setting regarding high-risk AI systems. In contrast, Korea’s current regulatory sandbox regime is characterized by a fragmented, sector-based structure operated by individual ministries and relies mainly on ex post regulatory exemptions and temporary approvals, which limits its capacity to address the cross-sectoral and high-risk nature of AI technologies. Based on this comparative analysis, this study argues that simply transplanting the EU model into the Korean legal system would be inappropriate. Instead, it proposes institutional adaptation, whereby the regulatory sandbox is redesigned to function as an integrated compliance infrastructure within Korea’s AI regulatory framework, and presents legislative and policy implications for improving the coherence and effectiveness of AI regulatory governance in Korea.