South Korea’s AI Regulation and the Future of Global AI Governance

South Korea’s Framework Act on the Development of Artificial Intelligence (AI) and Establishment of Trust came into force in January 2026, following its passage in January 2025. The Act makes South Korea the second major jurisdiction after the European Union to adopt comprehensive and binding AI regulation. This development takes place against a shifting global regulatory landscape. As the European Union pauses or revises parts of its AI regulatory regime following political and economic pressure from the Trump administration, South Korea is positioned to assume a leading role in global AI safety regulation.

The Act establishes a clear legal foundation for governing AI across development, deployment, and use. It adopts a risk based approach that focuses regulatory obligations on AI systems capable of causing significant harm. This approach prioritizes public safety, protection of fundamental rights, and institutional trust while allowing room for innovation and economic competitiveness.

The Act introduces the category of High Impact AI, defined as AI systems that have the potential to significantly affect human life, physical safety, or basic human rights. These systems include AI deployed in sensitive sectors as well as other areas that materially affect the protection of life, bodily integrity, and fundamental rights, as may be determined through Presidential Decree. This flexible structure allows the scope of regulation to expand as new high risk use cases emerge.

The Act also defines an Affected Person as any individual whose life, physical safety, or fundamental rights are significantly impacted by AI products or services. While this definition provides protection against physical and rights based harms, it does not explicitly include cognitive or mental harm. This omission is notable given the increasing reliance on generative AI systems that influence human decision making, reasoning and autonomy. Earlier technologies such as calculators, navigation systems, and search engines enabled users to offload discrete cognitive tasks. Generative AI differs in that it can perform complex reasoning processes on behalf of users. As dependency on these systems increases, there is a risk of reduced human decision making capacity and diminished autonomy. Current regulatory frameworks, including South Korea’s, largely address tangible harms while leaving these longer term cognitive impacts unregulated.

Article 3 of the Act outlines the basic principles and obligations of the state. It requires AI technology and the AI industry to be developed in a manner that promotes safety and reliability and contributes to improved quality of life. This principle establishes safety and trust as central objectives of national AI development.

The Act creates several institutional mechanisms to implement these objectives. The Minister of Science and ICT is required to establish an AI Basic Plan every three years to promote AI development, strengthen national competitiveness, and guide sector specific implementation. This plan is subject to review and approval by the National AI Committee, which is responsible for deliberating and resolving major AI policy issues and trust building measures.

A dedicated AI Policy Center is established to coordinate AI policy development and to support the formulation and dissemination of international AI norms. In addition, an AI Safety Research Institute is created to conduct specialized work aimed at protecting human life, physical well being, and property from risks associated with AI systems, while maintaining public trust in AI enabled environments.

The Act also promotes the standardization of AI technologies to support consistency, safety, and interoperability across sectors. Standardization is intended to reduce technical fragmentation and embed safety requirements into AI systems from the design stage.

Support for small and medium sized enterprises (SMEs) is explicitly included in the framework. When implementing support policies under the Act, SMEs must be given priority consideration, and the government is required to encourage their participation in the AI industry. This approach is particularly relevant for African countries considering AI regulation. SMEs account for the majority of employment across the continent and play a central role in economic growth. AI regulatory frameworks that fail to include SMEs risk excluding large segments of the private sector from digital transformation and innovation.

The High Impact AI classification carries specific compliance obligations. Businesses deploying AI systems are required to assess in advance whether their systems qualify as High Impact AI. Where classification is unclear, businesses may request confirmation from the Minister of Science and ICT. This mechanism provides regulatory certainty and reduces inconsistent self assessment.

Once an AI system is classified as high impact, businesses must comply with detailed operational requirements. These include the establishment and operation of a risk management plan, implementation of measures to explain AI generated outputs, and disclosure of key criteria used in generating those outputs, including a general overview of the training and operational data. Businesses must also ensure human management and supervision of high impact AI systems and prepare and retain documentation demonstrating compliance with safety and reliability obligations.

The requirement for human oversight is particularly significant given the increasing autonomy of AI systems. Mandatory human supervision helps ensure accountability, enables intervention where risks materialize, and limits over reliance on automated decision making in contexts with serious consequences.

South Korea’s Framework Act is likely to influence how countries approach AI regulation in the coming years. Its structured risk based model, clear institutional arrangements, and enforceable compliance obligations provide a practical reference point for jurisdictions seeking to regulate AI without undermining innovation. While gaps remain, particularly in addressing cognitive and mental harms, the Act represents one of the most comprehensive national efforts to align AI development with safety, accountability, and public trust.