The law library · Republic of Korea · national AI statute
South Korea: the AI Framework Act
Framework Act on the Development of Artificial Intelligence and Establishment of Trust, the “AI Basic Act” · in force with its Enforcement Decree on 22 January 2026
Page verified 5 August 2026. Source: official summaries, the published English translation of the Act, and contemporaneous legal analysis. Thresholds live in the Enforcement Decree and are not carried here.
The second comprehensive AI law in the world, and the first in Asia. It matters here for a narrow reason: alongside the EU and China, Korea has now put labelling of generative AI output into primary legislation. Three of the largest economies in the world, three different legal traditions, same requirement.
What it does
- Two purposesThe Act is explicitly a development statute as well as a trust statute. Its full title is the Framework Act on the Development of Artificial Intelligence and Establishment of Trust.
- High-impact AISpecific obligations attach to AI in critical sectors, named as including healthcare, energy and public services.
- LabellingMandatory labelling requirements for certain applications of generative AI.
- BaselineGeneral obligations for trust, safety, transparency and user protection across the board.
The pattern it completes
Set the three comprehensive regimes side by side and the convergence is hard to miss:
- EUAI Act Article 50. Machine-readable marking of synthetic content, applicable 2 August 2026, with systems placed on the market before that date given until 2 December 2026 by Regulation (EU) 2026/1744.
- ChinaLabelling Measures. Explicit and implicit labels, in force 1 September 2025.
- KoreaAI Basic Act. Mandatory labelling for certain generative AI, in force 22 January 2026.
Three legislatures with very different attitudes to regulation arrived independently at the same conclusion within eleven months of each other. That is what a durable requirement looks like as it forms, and it is a stronger signal than any single statute.
What changed since
The Enforcement Decree was amended with effect from 21 July 2026, adding eleven categories of AI-vulnerable persons to the high-impact AI impact-assessment regime. MSIT’s enforcement grace period runs about one year from 22 January 2026, so it lapses around 22 January 2027. The Korean government sites would not respond to automated retrieval, so we publish the dates and the substance and not a decree number.
What this page carries
Thresholds, the domestic-representative rule and the detail of the labelling duty live in the Enforcement Decree. This page carries the Act, its dates and its substance, and no article numbers or thresholds. Anyone with a Korean exposure should have the Act and the Decree read properly before acting on a summary, including this one.
What to do about it
This is an estate problem, not a document problem.
High-impact AI in healthcare, energy or public services is judged on what the system actually did on a given day, not on what the framework said it would do. Regulayer records the decision as it is taken, outside the system being governed, content-free, and verifiable by anyone holding the file with public mathematics, without a call to us.
Regulayer for enterpriseThe AI audit trail →
For an individual practitioner, the same engine is $349 a month with a free week: start sealing. See also the law library and every jurisdiction we track.
Sources
- Framework Act on the Development of Artificial Intelligence and Establishment of Trust, English translation, Center for Security and Emerging Technology
- Effective date confirmed across official and professional sources as 22 January 2026, together with the Enforcement Decree.
- Enforcement Decree not read. No thresholds or article numbers are given on this page.
Related
- The law library
- South Korea · PIPA
- China · AI content labelling
- AI rules for lawyers, country by country
Checked 5 August 2026. Information, not legal advice, and not advice on Korean law.
Common questions
When did South Korea's AI law take effect?
The AI Basic Act and its Enforcement Decree took effect on 22 January 2026, making South Korea the second jurisdiction after the European Union with a comprehensive AI regulatory regime, and the first in Asia.
What does the Korean AI Basic Act require?
It combines industrial promotion with baseline obligations for trust, safety, transparency and user protection. It creates specific duties for 'high-impact' AI systems in critical sectors such as healthcare, energy and public services, and imposes mandatory labelling requirements for certain applications of generative AI.
Does the Korean AI Act apply to foreign companies?
It is drafted to reach AI activity affecting the Korean market, and foreign providers above certain thresholds have obligations including, in some cases, appointing a domestic representative. Thresholds live in the Enforcement Decree. Read the Act and the Decree before acting on that summary.
How does Korea's law compare with the EU AI Act?
Both are comprehensive and risk-tiered, but Korea's frames itself around promoting development as well as establishing trust, and its 'high-impact' category is drawn by sector rather than by the EU's longer annexed list of high-risk use cases. Korea's regime became applicable on 22 January 2026. The EU AI Act became applicable on 2 August 2026, and Regulation (EU) 2026/1744 then moved its Chapter III high-risk obligations to 2 December 2027 and 2 August 2028. The two regimes are on different timetables.
Information, not legal advice, and not advice on the law of any jurisdiction. Every entry is verified against the issuing body’s own document; where a source is reporting rather than the document, we say so.