On January 22, 2026, South Korea’s AI Basic Act enters into force. Together with the final revisions the National Assembly passed in the closing days of December 2025, it becomes the first comprehensive national AI law anywhere to fully take effect — months before the EU AI Act’s full application in August 2026. With two weeks to go, any team whose product touches the Korean market should be checking whether they are in scope.
Officially the Framework Act on the Development of AI and the Creation of a Foundation for Trust, the law was adopted by the National Assembly on December 26, 2024, promulgated in January 2025, and merges 19 separate AI bills into a single statute. Korean coverage in the first days of January noted that the last-minute revision tightened risk controls while leaning further into industrial promotion.
A Law That Reaches Beyond Korea’s Borders
The scope is not limited to Korean territory. Any AI activity affecting the South Korean market — domestic or overseas — falls under jurisdiction, with an exception only for national defense and security; providers with no presence in Korea must appoint a domestic representative. The revision also lets universities and private companies, not just the science ministry, establish AI research institutes to build sovereign AI capability, and it requires public institutions to prioritize AI products and services in procurement. Governance gets an upgrade too: the Presidential Council on National AI Strategy gains formal legal status and expanded powers as the central coordinating body — coordinating ministries, setting major policy, and monitoring implementation in place of non-binding guidelines.
High-Impact AI, Defined by Compute
The revised act supplies an unusually concrete threshold: models trained with at least one hundred septillion (10^26) floating-point operations count as high-impact AI, then get assessed by application area, risk to fundamental rights, and severity. Examples include healthcare, hiring, loan screening, and biometric analysis — meaning a health-assistant feature like ChatGPT Health, if offered in Korea, would sit inside the review perimeter. Operators must maintain risk-management plans and continuously monitor social impact, and regulators can demand data submissions and conduct on-site inspections.
Labeling and Watermarks for Generative AI
The duties do not stop at high-impact systems. Every commercial AI service provider must disclose AI-generated content through notices or technical measures such as watermarks, aiming squarely at deepfakes, misinformation, and deceptive advertising. The practical read for product teams: build labeling into the interface and the output pipeline from the start. It is not a decoration to bolt on after launch.
Light Penalties, Long Grace Period
Violations — failing to label content, failing to appoint a domestic representative — carry fines up to KRW 30 million (about $21,000), with imprisonment possible under some provisions. Next to the EU AI Act’s turnover-based penalties, that is mild; analysts have noted the ceiling would not be meaningful for mid-sized and large companies. And for at least one year after the effective date, enforcement is guidance-first: officials will favor guidance, discounts on safety and reliability certifications, and expert consulting over punitive sanctions.
Industry Worries, and a Checklist
The nonprofit Startup Alliance warned that the fuzzy high-impact definition could land hardest on early-stage startups: if routine tools like machine translation or photo editing trigger watermarking duties, compliance costs will chill experimentation, investment, and market entry. It urged obligations to be phased in by company size and deployment risk rather than raw compute. Critics in the opposite direction argue the law tilts toward industrial promotion over enforceable protections for individuals — civil servants deploying AI are shielded from personal liability absent intent or gross negligence — and question enforcement while the implementing decrees are still being revised.
For developers and product teams, the practical checklist: inventory which features reach Korean users, check whether training compute crosses the 10^26 FLOP threshold, add content labels to generative outputs, line up a domestic representative if you have no Korean entity, and track the enforcement decree as it lands.
Sources
- How newly revised AI Basic Act will reshape Korea’s AI landscape — The Korea Times
- South Korea’s New AI law: What it Means for Organizations and How to Prepare — OneTrust
- South Korean AI Basic Law — Artificial Intelligence Act
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.