South Korea's AI Basic Act puts another AI governance regulation on the map

The article reports that South Korea has a pending law on AI, called the AI Basic Act, expected to take effect as law before the end of January. Overseas AI providers would need an in-country representative so their products meet safety and governance duties. Commentators quoted see some EU AI Act influence in the risk classification, human oversight, record-keeping and transparency, but view the Korean text as less prescriptive, leaving much to future decrees.

The legislative path is described in detail. A first attempt came in the 21st National Assembly, which ran from 2021 to 2023, and the bill was introduced again in the current one, merging over a dozen proposals. In late December 2024, 260 of 264 lawmakers voted for it. It reached the acting president on 10 January, and he has 15 days to sign or veto. Enforcement is set for January 2026, and one source expects no veto because all parties back the law.

On substance, high-impact AI, for instance in hiring, health care or essential services, requires advance user notice, risk management plans and human oversight. Large-scale AI needs a risk management system. Generative AI output must be labeled, with enforcement details left to later rules. Foreign providers above user or revenue thresholds, which a presidential decree will set, and without a Korean address must appoint a domestic representative handling safety reports and confirming high-impact status. Risk assessments are encouraged, not mandated. Where the Personal Information Protection Act overlaps, the more stringent rule governs. The act also promotes the industry via data centers, talent and data access. One expert sees no Brussels effect and expects industry-friendly guidance.