KLRI : KOREA LEGISLATION RESEARCH INSTITUTE

Ⅰ. Purpose of Research

○ Artificial Intelligence (AI), as a general-purpose and foundational technology driving next-generation innovation and growth, is rapidly permeating all sectors of society. Consequently, the deployment of AI across various societal domains and the corresponding policy responses have become critical priorities. Since South Korea announced its National AI Strategy in 2019, the country has actively pursued policies aimed at advancing AI technologies and fostering the AI industry. With the anticipated passage of the European Union’s Artificial Intelligence Act (AI Act) in 2023, the formalization of AI-related laws and regulations has become a major focus of global discourse. According to the 2023 Global AI Legislation Tracker, nations worldwide are actively developing AI governance frameworks and advancing legislation to address the implications of AI advancements. The EU Artificial Intelligence Act was officially enacted by the European Parliament in May 2024 and took effect in August 2024. Similarly, the United States issued an executive order in 2023 emphasizing the safe use of AI technologies, while the United Kingdom, Canada, and Australia published regulatory white papers on AI. As the global push for AI legislation accelerates, analytical research on AI legislative data has gained substantial attention. Stanford’s Institute for Human-Centered Artificial Intelligence (HAI), a globally renowned institution for assessing AI development metrics across nations, has identified critical components of its AI Leadership Index. These components include the prevalence and count of legislative bills referencing the term "artificial intelligence" and the extent to which such bills have been enacted into law. According to Stanford’s analysis, South Korea ranked fifth globally for the number of AI-related bills enacted between 2016 and 2023, with a total of 11 laws passed during this period. Notably, in 2023 alone, South Korea tied for second place globally, successfully enacting three AI-related laws. 

  ○ South Korea's legislative activity in the field of artificial intelligence has experienced substantial growth. In 2018, only six statutes or regulations explicitly referenced "artificial intelligence." However, by October 7, 2024, this figure had increased to 64. Additionally, as of the same date, a total of 255 bills containing the terms "artificial intelligence" or "AI" had been introduced into the legislative process, including 191 bills from the 21st National Assembly (2020 - 2024) and 64 bills from the 22nd National Assembly (2024 - October 7, 2024). This legislative data reflects South Korea's policy priorities, societal demands, and its responses to critical developments. In 2024, a comprehensive analysis of AI-related legislative data was planned to examine the applications of AI technology across various sectors and the associated challenges being addressed. This research aims to assess the current legislative landscape, identify areas requiring improvement in future AI-related legislation, highlight key issues, and propose actionable recommendations to enhance South Korea's legislative framework for artificial intelligence.

 

Ⅱ. Research Subjects

○ This research examines AI-related legislation proposed during South Korea's 21st and 22nd National Assemblies, as well as existing laws referencing artificial intelligence, as of October 7, 2024. Utilizing data from the Korean Law Information Center of the Ministry of Government Legislation, the study identifies and categorizes laws explicitly mentioning "artificial intelligence" in their provisions, summarizing their key contents and analyzing the legislative domains they address. Furthermore, leveraging the National Assembly Bill Information System, this research reviews bills introduced during the 21st and 22nd National Assemblies up to October 7, 2024, that include references to "artificial intelligence" in the legislative process. Among these, bills with medium to high relevance to AI technologies are selected for detailed analysis, with a focus on their primary content and legislative areas.

 

Ⅲ. Research Methods

▶ Overview

○ The legislative data analyzed in this research comprise 15 statutes and 49 subordinate regulations explicitly referencing "artificial intelligence," as well as 163 bills from the 21st National Assembly and 50 bills from the 22nd National Assembly that mention "artificial intelligence" during the legislative process and exhibit medium to high relevance to AI technologies. The analysis was conducted in four stages. First, the legislative data were classified and analyzed by policy areas, expansive or restrictive nature of legislative approach, and the AI technology lifecycle. Second, an analysis of key issues and common issues was conducted. Key issues included the regulation of high-risk AI applications and the legislative status of generative AI ― both globally recognized as critical topics in AI legislation. Common issues were identified from the Korean legislative data, revealing four recurring themes: the Framework Act on AI, Data, AI ethics, and AI-based administration (AI in digital government). Each theme was analyzed in detail. Third, a comparative legal analysis was conducted by reviewing and contrasting Korea's AI-related legislative data with that of leading AI regions, such as the United States and the European Union. Finally, this research summarized its findings and provided recommendations for addressing areas requiring improvement in Korea's future AI-related legislation.

▶ Legislative Data Analysis

○ This research not only provided a comprehensive overview of AI-related legislative data but also conducted a foundational classification based on the 44 legislative categories defined by the Ministry of Government Legislation on South Korea's legal information platform. Among the 15 statutes explicitly referencing "artificial intelligence" in their provisions, the distribution across legislative fields was as follows: ① General Administration (3 laws), ② Commerce, Trade, and Industry (3 laws), ③ Education and Academic Affairs (2 laws), and ④ Health and Medicine, Land Transport, Aviation, and Tourism, Elections and Political Parties, Information and Communications, Maritime Affairs, Social Welfare, and Pharmacy (1 law each). For the 49 subordinate regulations referencing "artificial intelligence," the primary legislative fields were identified as: ① Science and Technology, ② Commerce, Trade, and Industry, ③ Education and Academic Affairs, ④ Health and Medicine, and ⑤ General Administration.

○ The analysis of AI-related legislative bills identified the most active legislative fields as follows: ① Science and Technology, ② Commerce, Trade, and Industry, ③ Information and Communication, ④ General Administration, ⑤ Domestic Taxation, ⑥ Education and Academic Affairs, and ⑦ Culture and Public Information. A field-specific analysis of the legislative data revealed that the 21st National Assembly primarily focused on bills within the Commerce, Trade, and Industry sector, particularly those aimed at fostering the AI industry. Notably, despite being less than six months into its term, the 22nd National Assembly has already introduced bills in five legislative fields that were not addressed during the 21st National Assembly. As of October 7, 2024, a total of 32 legislative fields have accumulated AI-related legislative data based on bills alone. When statutes and subordinate regulations are included, this number expands to 35 legislative fields. These findings highlight the rapid expansion and diversification of AI-related legislative efforts in South Korea.

▶ Analysis of Legislative Status

○ Using the compiled legislative field data, an analysis of the legislative status was conducted in three stages, incorporating the results of a national public survey. First, the legislative status by policy area was examined to identify gaps relative to public policy demands. According to the national survey on major policy areas in Korea, the public ranked the need for AI technology adoption in policy areas as follows: ① Transportation, ② Taxation and Finance, ③ Legislative, Judicial, and Executive Services, ④ Public Resources, ⑤ Health and Welfare, ⑥ Safety and Disaster Management, ⑦ Education, ⑧ National Defense, ⑨ Culture and Arts, ⑩ Labor and Employment, ⑪ Agriculture, ⑫ Environment, ⑬ Finance, ⑭ Land and Buildings, and ⑮ Energy. A comparative analysis of the proportion of Korea's AI-related legislative data against public demand rankings revealed notable gaps in the following fields: ① Transportation, ② Safety and Disaster Management, ③ Environment, and ④ Agriculture. Furthermore, while the public has expressed a strong demand for AI adoption in administrative services to enhance accessibility, fairness, transparency, and convenience, existing legislative data in taxation and finance, as well as legislative, judicial, and executive services, primarily focus on tax benefits to support the AI industry and stricter penalties for crimes involving deepfakes. These findings underscore the need for additional legislative measures to establish a legal foundation for the adoption of AI technologies in administrative services.

○ As of October 7, 2024, an analysis of South Korea’s legislative data revealed a clear trend favoring expansive legislation, with 119 expansive bills compared to 72 restrictive ones. Notably, 22 bills were categorized as having a complex nature. A comparison of the legislative tendencies between the 21st and 22nd National Assemblies highlights this shift. The 21st National Assembly introduced 101 expansive bills and 51 restrictive ones, resulting in nearly twice as many expansive bills. This pattern aligns with the legislative priorities set forth in the 2019 National AI Strategy, which focused on fostering a supportive environment for this research, development, and industrial foundation of AI technologies. In contrast, the legislative activity of the 22nd National Assembly reveals a shift in focus. Of the bills introduced, 18 were expansive, while 21 were restrictive, signaling an increased emphasis on addressing the potential misuse and societal impacts of AI technologies. This indicates a transition in legislative priorities, as post-2024 legislation increasingly incorporates safeguards and regulatory measures to mitigate the risks associated with AI's rapid adoption and integration into society.

○ In the third stage of analysis, legislative data were categorized according to the lifecycle of artificial intelligence, focusing on its utilization in society. The lifecycle was divided into five stages: ① Establishing an industrial foundation for this research and development of AI technologies, ② Developing AI algorithms, ③ Collecting, storing, analyzing, and utilizing data, ④ Applying AI technologies to specific industries, and ⑤ Addressing the impacts of AI technologies on end-users. Among AI-related bills in South Korea, the largest number fell under the 5th stage (67 bills), which focused on mitigating risks and addressing societal issues arising from the end-user consumption of AI technologies. The second most common category was the 1st stage (50 bills), aimed at establishing the foundational infrastructure for AI research, development, and industrial growth. This was followed by the 4th stage (46 bills), which supported the integration of AI technologies into various industries. The 3rd stage (20 bills) pertained to legislation governing the collection, storage, analysis, and use of data, while the least represented category was the 2nd stage (8 bills), focusing on regulating the algorithms used in AI applications.

▶ Analysis of Legislative Trends by Key Issues

○ The first key issue in AI legislation and regulation is high-risk AI. Globally, legal frameworks for advanced technologies adopt a risk management framework to control the risks associated with these technologies. In the case of AI, laws specifically identify and regulate high-risk areas of AI use. In South Korea, most bills resembling a Framework Act on AI incorporate provisions for designating and regulating high-risk AI areas. These bills commonly classify AI applications in social infrastructure, public sectors, healthcare, recruitment, and emergency disaster response as high-risk. A notable feature, however, is that all bills exclude education and judicial sectors from high-risk classifications. The 2024 national public survey results confirmed that the public supports the need to regulate high-risk AI areas through legislation. However, respondents indicated that AI applications in certain areas do not require classification as high-risk, such as AI in admission decisions, grading, and ranking in educational institutions (52.6% deemed not risky); AI assisting in factual and legal judgments in court (52.4% deemed not risky); and AI used in lie detectors or evaluating the reliability of evidence during investigations (51.3% deemed not risky). In contrast, respondents assessed AI as risky in managing and operating social infrastructure and energy systems (58.5% deemed risky), healthcare and medical devices (53.8% deemed risky), and emergency disaster response (47.6% deemed risky). By integrating legislative data with survey results, it becomes evident that legislative measures to regulate high-risk AI are essential. Furthermore, areas such as emergency disaster response and energy system management and operation require additional legislative attention and enhancement to address identified risks effectively.

○ The second key issue pertains to generative AI. In 2023, legal challenges associated with generative AI models, such as Chat GPT―capable of producing diverse unstructured content based on large language models―began to surface. Major concerns regarding generative AI include the development and deployment of technologies for identifying AI-generated content, accountability for their implementation, measures to prevent misuse and abuse, penalties for violations, and copyright implications. Among the 52 bills analyzed from the 22nd National Assembly, 17 specifically target the prevention of misuse and abuse of generative AI and deepfake technologies, as well as the provision of remedies for damages caused by these technologies. This indicates that there is currently a significant body of legislative data focusing on the regulation of AI technologies linked to digital crimes, such as deepfakes. However, legislative data explicitly addressing obligations like watermarking requirements for products and services utilizing generative AI, as well as resolving copyright issues related to the outputs of generative AI models, remain insufficient.

○ The legislative data were analyzed to identify four common issues: the Framework Act on AI, Data, AI Ethics, and AI-based Administration. As of October 7, 2024, 11 bills related to the Framework Act on AI were proposed during the 21st National Assembly (2020 - 2024), with an additional 11 bills introduced in the 22nd National Assembly. A comparative analysis was conducted on the structure and content of all bills related to the Framework Act on AI proposed in South Korea’s National Assembly. Secondly, legislation related to AI and data was examined to assess its scope and focus. Thirdly, legislative efforts addressing AI ethics principles―such as safeguarding human rights, respecting diversity, promoting the public interest, protecting privacy, preventing harm, ensuring accountability, and guaranteeing safety―were analyzed. Lastly, laws concerning AI-based administration were reviewed to evaluate their development and implications for governance.

▶ Comparative Legal Research

○ An analysis of AI-related legislative data from the United States and the European Union reveals that The United States possesses a more comprehensive repository of legislative resources, particularly in the domains of military and defense, monetary policy and fiscal governance, and international relations, which are not extensively developed within the South Korean legislative framework. While the United States does not typically regulate high-risk AI systems through comprehensive legislation, specific bills have been proposed to address targeted concerns. These inclu de restricting the continuous surveillance of workers using AI in the workplace and overseeing AI applications in fields such as telemedicine, firefighting, and defense-related nuclear activities. Unlike South Korea, the United States has not enacted a comprehensive Framework Act on AI. Instead, it addresses AI-related issues through executive orders and sector-specific legislation. This approach reflects a regulatory framework characterized by a sector-specific, incremental strategy, often referred to as a "bottom-up" model.

○ In contrast, the European Union (EU) has adopted a comprehensive and horizontal regulatory framework for artificial intelligence, the EU AI Act, which applies broadly across all sectors of society. As a result, the EU has fewer sector-specific legislative initiatives compared to the United States. However, the EU has established a wide array of legal safeguards aimed at protecting individuals, learners, consumers, and workers in areas such as fundamental rights (as enshrined in the EU Charter of Fundamental Rights), environmental sustainability, and education. These protections underscore the EU’s commitment to ensuring that AI deployment aligns with its overarching principles of human dignity, social welfare, and environmental stewardship. 

 

Ⅳ. Findings and Recommendations

○ Based on the findings of this research, it is concluded that South Korea  has established a multifaceted legislative framework for the utilization of artificial intelligence (AI). However, despite increasing public demand for the strategic application of AI technologies in key areas such as (1) transportation, (2) safety and disaster management, and (3) environmental protection, legislative efforts in these domains remain insufficient. To address these gaps, it is essential to enhance the legal framework in the transportation sector by fostering diverse applications of AI technologies that allow citizens to experience their benefits in daily life. Additionally, in the safety, disaster management, and environmental sectors, legal measures must be established to not only promote the adoption of AI technologies but also mitigate potential risks associated with their use.

○ Regarding the designation of high-risk AI domains, a critical issue in the legal and institutional regulation of artificial intelligence technologies, there is strong demand for government oversight and management in high-risk areas. However, legislative proposals concerning high-risk AI consistently exclude the education and judicial sectors, and public opinion surveys similarly indicate that the perceived risks of applying AI technologies in these fields are not significant, which reflects the unique context of South Korea. Nevertheless, high-risk domains should be designated with careful consideration of the actual risks and impacts on citizens' fundamental rights. Conversely, both legislative data and public surveys highlight insufficient legislation addressing AI systems used in emergency disaster response and energy management and operation. To mitigate the unpredictable risks associated with AI technologies, the government must implement targeted management measures for high-risk AI domains. These domains, however, should be defined carefully, ensuring alignment with the policies and priorities of both the South Korean government and its citizens. Regarding generative AI, there is an urgent need to strengthen legislation in key areas, including mandatory watermarking for user recognition, the allocation of liability, and copyright issues related to its application.

○ Understanding the legislative trends of leading nations in artificial intelligence is critical for predicting the trajectory of technological advancement and the core values that underpin AI development in those countries. In South Korea, the most pressing issue in AI-related legislation is the need to establish a comprehensive Framework Act on AI. Concurrently, the government is actively promoting policies and legislation to enhance the application of AI technologies in public administration. From a legislative perspective, one-third of the total legislative data pertains to bills in the science and technology sector and the commerce, trade, and industry sector, both of which are pivotal for fostering the development and growth of AI technologies. This reflects South Korea’s emphasis on legislation that promotes industrial growth and supports AI innovation. By comparison, the United States focuses its legislative efforts on maintaining its status as a global AI leader and advancing national security objectives. The European Union (EU), meanwhile, has implemented a comprehensive AI framework law and continues to pursue legislation that safeguards environmental and constitutional values for its citizens. Drawing on examples from these leading AI jurisdictions, U.S. legislative cases offer valuable insights for Korea in drafting bills to protect national AI technologies, particularly in areas requiring enhanced security measures to ensure global competitiveness. Similarly, EU legislative models provide guidance for addressing societal risks and mitigating the negative impacts of AI technologies, including public safety, health, and the protection of fundamental rights.