This is what the South Korea AI Basic Act changed. South Korea’s Framework Act on the Development of Artificial Intelligence and Establishment of Trust, known as the AI Basic Act, came into force on 22 January 2026. Its Enforcement Decree took effect the same day. It is the first comprehensive national AI law in Asia and, after the EU AI Act, one of the first anywhere.
What South Korea’s AI Basic Act requires
The Act places duties on AI business operators, the companies that develop AI or use it in their products or services:
- Labels on generative AI output: businesses offering generative AI must tell users in advance that a product or service uses it, and mark its output as AI-generated.
- Clear labels on deepfakes: realistic synthetic images, audio or video that could be mistaken for real must be labelled so people can tell.
- Duties for “high-impact” AI: AI systems used in sensitive areas such as health care, hiring, lending, energy and public services need risk management, explanations of how they reach results, human oversight and record-keeping.
- Extra AI safety duties for the largest models: very large advanced AI systems, such as those trained with more than 1026 FLOPs of computing power, must manage AI risk by identifying, assessing and reducing it.
- A local representative: large foreign AI providers above thresholds set by decree must appoint a representative in Korea. A provider qualifies with, for example, KRW 1 trillion in total revenue, KRW 10 billion in AI-service revenue, or one million average daily users in Korea.
Enforcement and fines
The Ministry of Science and ICT oversees the AI Basic Act and can investigate and order corrections. Fines for breaches such as failing to tell users in advance about AI use, failing to appoint a representative or ignoring a correction order are capped at 30 million won, much lower than EU penalties. The government has said it will focus on guidance during a grace period of a year or more before imposing fines, except in cases of serious harm.
The ministry describes its approach as “minimum regulation”. It has issued guidelines on high-impact AI, impact assessments, transparency for generative AI and the safety of advanced AI, and a help desk run by the Korea Software Industry Association answers compliance questions.
Why it matters for AI governance
On AI regulation, Korea’s new AI law sits between the EU and Japan. Like the EU, it seeks trustworthy AI: it regulates high-impact AI and requires labels. Like Japan, it also aims to grow the AI industry and the development of AI technology, and its penalties are modest. For companies serving Korean users, the practical effect is clear: label generative AI and deepfakes, and document how high-impact AI systems are managed as part of your AI governance.
What to do now
- If your app or service uses generative AI in Korea: tell users up front and label output, especially realistic images and audio.
- If you are a large foreign AI provider: check whether you meet the thresholds for a local representative.
- If you use AI to make decisions in health, finance or hiring: prepare risk management and explanation processes.
How it compares with the EU AI Act and Japan
- EU AI Act: sorts AI by risk, bans some uses outright and imposes detailed duties on high-risk AI systems, with fines up to 7% of global turnover.
- Korea’s AI Basic Act: regulates high-impact AI and requires labels, but has no list of banned uses and much lower fines. It also commits the government to supporting the AI industry.
- Japan’s AI Promotion Act: promotes AI innovation and coordinates AI policies, with no fines at all.
South Korea AI Basic Act: common questions
Does the AI Basic Act apply to foreign companies?
Yes. It applies to activities carried out abroad that affect the Korean market or Korean users, which is why large foreign providers may need a local representative.
Do I have to label every AI-generated image?
The Act requires generative AI output to be marked, and realistic deepfakes to be labelled in a way people can clearly recognise. The enforcement decree and guidance set out exactly how, including when a machine-readable mark is enough and when a visible notice is needed. Output that stays inside a service, such as a chatbot or game, can use on-screen notices or logos, while content people can download should carry clearer labels.
Does it replace Korea’s privacy law?
No. The AI Basic Act sits alongside the Personal Information Protection Act. If an AI system uses personal data, the privacy rules still apply.
Detailed rules come through the Act’s enforcement decree and ministry guidance, which can change. Our South Korea tracker entry has the latest status and sources.
Last checked: 5 November 2026. This explains the law in plain English; it is not legal advice. Spotted an error? Tell us, and see how we check facts.
