Korea's AI Basic Act Is Live: The Rulebook Every Global AI Company Quietly Has to Read in 2026
While the world argues about AI laws, South Korea simply switched one on. Here is what it says, who it reaches, and why a chatbot company in California may be affected.
Here is a question worth asking this week: if your app uses AI and has a million daily users in Korea, do you have a legal contact in Seoul?
Since January 22, 2026, South Korea's AI Basic Act has been in effect. It is a national framework law for artificial intelligence, and its reach goes beyond Korean companies. Foreign AI services that affect Korean users can fall under it too. We are now several months into the rollout, and the part everyone should be watching is the one-year grace period before fines begin.
What the Law Actually Covers
The Act splits companies into two groups: those that develop AI, and those that offer products or services that use AI. Both are covered. The definition of AI is broad, which means the law is designed to age well rather than be rewritten every time a new model appears.
The heavy obligations land on what the law calls high-impact AI: systems that significantly affect human life, safety or fundamental rights. Think healthcare, energy, transportation, hiring, and biometric analysis.
1. Assess whether your system qualifies before deployment
2. Provide a meaningful explanation of outcomes
3. Summarize the training data
4. Create a user protection plan
5. Keep human oversight and documentation
I have watched a lot of AI regulation announcements turn into thick PDFs nobody follows. What stands out here is the "meaningful explanation" requirement. Anyone who has been rejected by an automated hiring screen knows how rare a real explanation is. If this is enforced well, that single clause could matter more than any fine.
The Part That Reaches Overseas
The Act applies to AI systems operated outside Korea if they affect Korean users or the Korean market, with exceptions for national security uses. Foreign businesses with no Korean office must appoint a local representative if they cross any one of three thresholds.
| Threshold (prior year) | Amount |
|---|---|
| Total company revenue | Over 1 trillion won (about US$681 million) |
| AI service revenue | Over 10 billion won |
| Average daily users in Korea | Over 1 million in the three months before year-end |
That representative handles government inquiries and safety reports. Skipping it is one of the violations that can draw the maximum 30 million won fine.
The fine is small by tech standards. Roughly US$21,000 will not scare a global platform. But the law also lets the government order a service to be suspended if it poses a safety threat. The real lever is access to a market that is among the most connected in the world, not the penalty.
Labels on AI-Made Content
If generative AI produces sound, images or video that are hard to tell apart from human-made content, users must be clearly told. Products and services built with generative or high-impact AI must also give advance notice that AI is involved. For creators and K-content studios using AI tools, that translates to labels, not bans.
Insider's Insight: why this matters for K-culture
Korea exports webtoons, dramas and music at scale, and AI tools are now part of that pipeline. A clear labeling rule at home sets expectations for audiences abroad, and studios that build the habit early will have fewer surprises later.
The Government's Side of the Bargain
The Act is not only a rulebook. It sets up a National AI Committee chaired by the president, described as a "control tower" for national AI policy, and creates an AI Policy Center and an AI Safety Research Institute. It also requires support for research and development, data centers, small businesses and startups. Systems trained with cumulative compute of at least 10^26 FLOPs are designated high-performance AI, with added duties like risk management plans and reporting to the Ministry of Science and ICT (MSIT).
The one-year grace period is MSIT's stated intention, not a statute you can cite in court. Details of enforcement decrees can change, so if your business is affected, read the current guidance from the ministry or talk to a lawyer. I am summarizing a law firm overview, not giving legal advice.
Key Takeaways
- Korea's AI Basic Act took effect on January 22, 2026, covering both AI developers and companies that use AI in their services.
- Foreign firms above revenue or user thresholds must appoint a local representative, and fines reach up to 30 million won.
- Expect labels on AI-generated media and explanations for high-impact AI decisions.
Final Thoughts
Korea has bet that clear rules and heavy investment can coexist. Whether that bet pays off depends on enforcement over the next year. For now, the signal to the world is simple: build AI that Korean users rely on, and be ready to explain it.
Your turn: should every country require AI-generated images to be labeled, or does that go too far? Share your view in the comments.