인공지능기반행정의 알고리즘 규제에 관한 법적 고찰

Abstract With the recent emergence of AI, machines are evolving beyond merely assisting human judgment to the point of replacing it altogether. Various forms of AI are being utilized across all administrative areas now. This is characterized by enhanced speed and efficiency of administration as the AI recognizes and learns patterns autonomously and automatically carries out administrative tasks. Additionally, it also enables a proactive and effective response to natural disasters by analyzing past data to predict the future or by recognizing potential risks in advance. With the advancement of AI technology, AI-driven public administration is gradually being adopted by countries. What fundamentally differentiates AI from existing technologies or machines is its capacity for learning and autonomy. In addition, the technical characteristics of AI are viewed as an evolutionary process from automation to autonomy. The two are distinguished based on the presence or absence of human control. Autonomy refers to actions being controlled solely by the system without human intervention. The problem is that learning capacity and autonomy are difficult to reconcile with the characteristics of administration. AI-driven public administration is an administration based on AI and can be considered a typical type of fully automated administrative decision-making. The concept of fully automated administrative decision-making has been examined within the same or overlapping context as electronic administration, administrative acts conducted through automated machinery and automated administrative decision-making previously discussed. However, it presupposes a more advanced and comprehensive level of automation than these earlier concepts. Article 20 of the General Act on Public Administration is the legal basis for AI-driven public administration. It is generally viewed that the inclusion of AI in Article 20 of the General Act on Public Administration is an active administrative response to the emerging phenomenon of AI. this study examines algorithm regulation in AI-driven public administration through a comparative analysis of major countries, categorizing it into four key types : (1) regulation for personal information protection, (2) regulation for ensuring transparency, (3) regulation on the level of responsibility in case of harm caused by algorithms and (4) issues related to the application of algorithms in the public administration sector.