AI기본법 하위규정안의 개선방안과 해석론 -규제대상과 보호대상을 중심으로

The proposed rules for the AI Framework Act, which the Ministry of Science and ICT released on September 17, 2025, exclude deployers (users for official or business purposes) from the regulated “AI Use Business Operators” (UBOs) and instead incorporated them into the protected “users”. It aims at mitigating pan-industry shocks from regulatory duplication and overregulation inherent in horizontal regulatory frameworks. However, a failure to concurrently amend the statute has caused several side effects. First, combined with existing legislative errors, the AI Framework Act has been transformed into (i) regulation of AI-based product and service usage relationships rather than AI system regulation, and (ii) regulation where developers and distributors must protect deployers, who often have dominant positions, rather than affected citizens. Second, as deployers were taken out from covered entities, compliance items for the deployment process were assigned mostly to UBOs and partially to “AI Development Business Operators,” resulting in excessive obligations for both. The AI Framework Act lacks value chain obligation allocation concepts present in the EU AI Act, such as intended purpose and substantial modification, failing to avoid interpretation that upstream developers (particularly, general-purpose model developers) bear responsibility whenever high-impact use cases are added downstream beyond their foresight and control. Third, legal uncertainty has been created concerning whether the foregoing rulings align with the AI Framework Act. After further examining detailed issues in the proposed rules, this paper demonstrates the necessity of statute-level amendments for addressing these problems, deriving legislative directions and purposive interpretations to mitigate side effects before the completion of the amendments.