AI 알고리즘의 투명성과 설명요구권에 관한 비교법적 고찰 ‒EU AI Act와 미국 행정명령을 중심으로‒

As artificial intelligence (AI) technologies continue to expand, algorithmic decision-making systems are increasingly employed in areas such as public administration, finance, employment, and social welfare. While these systems enhance efficiency and consistency, their opacity raises significant legal concerns regarding individual rights protection and democratic accountability. In particular, where automated decisions have a substantial impact on an individual’s legal status or substantive opportunities, the ability to obtain an explanation becomes a crucial element of procedural fairness and effective remedies. This article examines the legal nature of “explanation” in AI-driven decision-making through a comparative analysis of the regulatory frameworks of the European Union, the United States, and Korea. The EU, through the AI Act, adopts a risk-based regulatory approach that imposes ex ante transparency and explanation-related obligations on AI providers and deployers. The United States, by contrast, relies primarily on executive orders and administrative governance, treating explainability as a policy objective and a component of accountability rather than as an enforceable individual right. Korea, however, represents a distinctive model by explicitly recognizing the right to request explanations for automated decisions under the Personal Information Protection Act and the Credit Information Act. This article argues that Korea’s rights-based model offers a meaningful mechanism for post hoc control of automated decision-making but also reveals structural limitations, particularly its fragmented legal basis and its reactive nature. Accordingly, the article suggests that the right to explanation should be preserved as a core procedural right while being complemented by ex ante risk management mechanisms inspired by comparative regulatory approaches. Through this analysis, the article seeks to contribute to the ongoing debate on balancing individual rights protection and institutional design in AI regulation.