一项法律制定了韩国的人工智能政策，一个薄弱环节可能会打破它（英）

SEJIN KIM AND HODAN OMAAR |SEPTEMBER 2025 By uniting strategy, promotion, and regulation in a single law, South Korea has given itself apowerful instrument to shape AI—but its blunt regulatory mandates threaten to drag down thevery strengths that make the act ambitious. KEY TAKEAWAYS South Korea’s AI Framework Act is the first in the world to combine strategy, industrialpromotion, and regulation into a single statute, magnifying both its strengths and itsflaws. South Korea’s promotion agenda—data infrastructure, clusters, talent, andinternationalization—is strong and forward leaning, putting it at the leading edge amongpeers. The act gets strategy and industrial policy broadly right, but it undermines them withblunt regulatory provisions that misdiagnose AI’s risks. Overly broad definitions, rigid R&D mandates, and SME-first rules risk stifling the scaleand flexibility South Korea needs for global competitiveness. Heavy-handed tools such as labeling requirements, compute thresholds, and process-heavy reporting will waste resources without delivering real accountability. Fixing the act requires two levels of adjustment: structural amendments by the NationalAssembly and balanced, practical implementation through MSIT’s Enforcement Decrees. CONTENTS Key Takeaways ................................................................................................................. 1Introduction ..................................................................................................................... 3Overview of the Act ........................................................................................................... 6Chapter 1: General Provisions ............................................................................................ 7Analysis: The Definition of an AI System Is Too Broad........................................................ 8Recommendation........................................................................................................... 9Chapter 2: Governance for Sound AI Development and Trust.................................................. 9Analysis: Keep the Master Plan, Lose the Master Regulator .............................................. 10Recommendation......................................................................................................... 10Chapter 3: Policies for AI Development and Industry Promotion ........................................... 10Innovation Polices to Support South Korea’s AI Industry .................................................. 10Analysis: South Korea’s Prescriptive Innovation Policies Risks Stifling AI ........................... 11Recommendations ....................................................................................................... 12Industrial Policies to Support South Korea’s AI Industry................................................... 12Analysis: Industrial Policy Measures Should Be Size-Neutral ............................................ 13Recommendations ....................................................................................................... 14Chapter 4: AI Ethics and Trustworthiness .......................................................................... 14Soft-Law, AI Safety-Oriented Measures........................................................................... 14Hard-Law, Regulatory Obligations for AI ......................................................................... 14Analysis: A Light-Touch Vision for AI Governance Undone by Heavy-Handed Rules.............. 15Recommendations ....................................................................................................... 17Chapter 5: Supplementary Provisions ................................................................................ 18Analysis: Broad Data Demands Risk Regulatory Overreach................................................ 18Recommendation......................................................................................................... 19Chapter 6: Penalties........................................................................................................ 19Analysis: Penalties Should Be Proportionate to Risk and Harm.......................................... 19Recommendations ....................................................................................................... 19Conclusion..................................................................................................................... 19Appendix: AI Act Summary .............................................................................................. 20Chapter 1: General Provisions (총칙).............................................................................. 20Chapter 2: Promotion System for the Sound Development and Trust-Based Foundationof AI (인공지능의건전한발전과신뢰기반조성을위한추진체계)................................... 20 Chapter 3: AI Technology Development and Industry Promotion (인공지능