미국 AI 행정명령의 라벨링에 관한 연구 = A Study on Labeling of the AI Executive Order in U.S.A.

다국어 초록 (Multilingual Abstract)

The United States Constitution explicitly establishes the separation of powers by granting the executive branch the power to execute laws. In general, this constitutional authority over administration allows the President to issue EO(Executive Order)s as means of directing federal agencies toward adequate measures for sectoral policies. However, typically EOs do not have direct binding force upon citizens as the legal effect on the general public. Given the need for administrative measures to address the problems related to AI(Artificial Intelligence) like deep-fakes’ issues, even if not enacted by Congress, the implementation of policy declarations and regulatory approaches can be considered.
Therefore the U.S. situation, where administrative actions based on AI EO accelerate initiative to develop and manage AI, provides a noteworthy points. But it is inevitably necessary for paying close attention to comparison as our domestic legislative system of discussions on AI unfold. Accordingly, the implications in administrative implementation schemes designed for enhancing the trustworthiness and transparency of artificially generated or manipulated content by adoption of AI technologies, are significant, which are ruled by AI EO regulating the so-called labeling in the capacity of AI content marks including another measures such as watermarking.
In our country, there has been a distinct approach in judicial precedents regarding administrative order distinguishing as administrative regulation or individual administrative disposition. Besides, there is no consistent normative or procedural framework for administrative ordinances like EOs within our constitutional system. As a consequence, regardless of how we conceptualize administrative ordinances, they do not carry the same authority and functions as EOs issued by the U.S. President. Nevertheless, when the government proposes enactment or amendment, pre-legislative scrutiny through review of administrative rule-making allows for a more deliberative process than direct legislative actions by lawmakers. In this context, effective labeling interlocked with content moderation in utilizing AI content would warrant trustworthiness and transparency for AI system in our legal framework.