South Korea AI Basic Act
Korea's framework AI law. It sets up national AI governance (a strategy committee, basic plan, AI safety institute) and industry support, and places duties on businesses: advance notice when products use high-impact or generative AI, labelling of generative and deepfake output, risk management for high-impact AI, and safety duties for very large frontier models.
At a glance
In force
- Jurisdiction
- South Korea
- Type
- Law
- Official title
- 인공지능 발전과 신뢰 기반 조성 등에 관한 기본법 (Framework Act on the Development of Artificial Intelligence and Establishment of a Foundation for Trust), Act No. 20676
What this means for you
If you build AI
- Generative AI output must be labelled; deepfake-like output must be clearly recognisable, while non-deepfake output such as animation may use invisible watermarks.
- Safety duties apply to models trained with at least 10^26 FLOPs that are cutting-edge and pose wide, serious risks to basic rights.
If you deploy AI
- Using AI in high-impact areas set by the law brings risk-management and user-protection duties; the decree treats a human making the final decision as controllable and outside high-impact.
- The ministry says fact-finding and fines are on hold for at least one year from 22 January 2026, except in extreme cases such as loss of life.
If you use AI
- You should be told in advance when a product or service runs on high-impact or generative AI.
- Watermarks or labels should help you spot AI-generated content and deepfakes in Korea.
Key dates
- Dec 26, 2024PastNational Assembly passes the AI Basic ActSource
- Jan 20, 2026PastAmendment (Act No. 21311) promulgated; some added provisions apply six months after promulgationSource
- Jan 22, 2026PastAI Basic Act and its Enforcement Decree take effect; fine grace period of at least one year beginsSource
Status history
In the Drops
No Drop has covered this law yet.