유네스코 ‘인공지능 윤리에 관한 권고’에 대한 법적 고찰

Regarding the rules of AI, guidelines or standards have been applied rather than norms with clear force. However, recently, there has been a trend towards enacting relatively clearer and more coercive laws. UNESCO adopted ‘the Recommendation on Ethics for Artificial Intelligence’(‘the Recommendation’) on 24 November 2021. This is an attempt to induce active cooperation among member countries by establishing international standards with more specific normative effect for AI. The Recommendation is the first global normative standard for AI and 193 member countries are obliged to respect this recommendation in establishing national norms on AI. This paper analyzes the main contents and characteristics of the Recommendation and summarizes the implications that should be considered when preparing a normative response to AI in South Korea. The Recommendation consists of a preamble, 8 chapters and 141 articles. Chapter 3, which presents the values and principles that AI actors must abide by and Chapter 4, which specifically specifies the measures to be taken by member countries, are particularly important. Its characteristics can be summarized in the following four points. First, it focuses on guaranteeing human-centered values ​​from a broader perspective, premised on human beings in the community living together, human beings prospering in the environment, ecosystem, and cultural diversity. Second, it is emphasized that all AI ethics can be realized only when they are based on trust, and that trust can be built through continuous monitoring and evaluation that allows for transparency, accountability, and participation of all stakeholders. Third, it presents specific policies that member countries must follow in each field by dividing the fields in detail along with the structural characteristics of being very systematic. Fourth, in relation to the regulation on AI systems, it value dialogue and consensus, as well as broad participation and cooperation of various stakeholders. The implications of the Recommendation can also be summarized in four points. First, it reminds us of the need to establish legal norms with a certain force to regulate AI. Active discipline by the state is essential for the establishment and sustainable development of an AI society based on trust. Second, it is necessary to think about the method and level of the regulation. While reflecting the main goals and principles of the Recommendation, it is necessary to differentiate the level and content of discipline by distinguishing whether or not it is an area or a problematic situation that should be directly regulated by the state. While acknowledging the potential for private autonomy to be exercised, it is also necessary to materialize support and coordination for this. Third, the normative and organizational system for the discipline of AI should be established. It is most reasonable to include matters to be specifically regulated in each area and issue in the individual acts, assuming that the basic act stipulates the standards and principles with only AI in mind, and this point of view also can be applied to the organization matter. Fourth, the norms should actively reflect flexible and proportional discipline, the participation and cooperation, and comprehensive education and support for this, taking the principle of human-centered development as a premise.