South Korea’s Ministry of Science and ICT (MSIT) issued a consolidated package of supplementary regulations to accompany the AI Basic Act (effective 22 January 2026) — the world’s second comprehensive AI law after the EU AI Act. The package includes: the Enforcement Decree (defining scope, institutional roles, and procedural requirements); high-impact AI definitions with sector-specific criteria; compute threshold rules (10²⁶ FLOPs); self-assessment frameworks for AI providers; and a ministerial confirmation pathway for borderline cases.
The regulations operationalise the Act’s obligations, clarifying which AI systems qualify as high-impact under the Act and how the National AI Committee, AI Policy Center, and AI Safety Research Institute will function. The Library of Congress confirmed the Act came into force on 22 January 2026.
Penalty provisions are deferred by one year (enforcement from January 2027), giving organisations an initial guidance-first period before financial consequences apply.
Who it affects: AI system providers and deployers operating in South Korea across all sectors. The high-impact AI definition with sectoral criteria directly shapes which organisations face transparency, documentation, and worker notification obligations under the Act.
What is notably missing: The penalty deferral means the enforcement regime is not yet fully active. The Act’s transparency provisions for workers (already scored) apply immediately, but no training-specific enforcement mechanism has been enacted — penalties are for AI system non-compliance, not for failure to train employees.
Secondary source: https://www.jdsupra.com/legalnews/south-korean-ministry-of-science-and-4483436/