South Korea's Ministry of Science and ICT began enforcing the high-impact provisions of the country's AI Basic Act this month, requiring Samsung Electronics, Naver, and Kakao to complete formal risk assessments for AI systems used in hiring, credit scoring, healthcare diagnostics, and other sensitive categories. The enforcement phase follows a nine-month grace period after the law took effect in January 2026, and it marks the first binding AI safety framework to reach full application anywhere in Asia.
What counts as "high-impact" under the law
The Act defines high-impact AI as systems that materially affect a person's life, safety, or fundamental rights, a category that in practice covers algorithmic hiring tools, medical imaging systems, autonomous driving software, and credit or insurance underwriting models. Companies operating systems in these categories must register them with the ministry, document training data sources, and run bias and safety testing before continued deployment is permitted. Non-compliance carries fines of up to 30 million won per violation, along with the possibility of a forced suspension order for systems found to pose unmanaged risk.
Samsung's Galaxy AI features, including on-device translation and photo-editing tools built into its Galaxy S26 lineup, fall largely outside the high-impact category, but the company's separate AI-assisted recruitment platform used for corporate hiring in South Korea does not. Samsung said it has already submitted registration documents for that system and expects the review to conclude before the end of the year.
Naver and Kakao face a heavier compliance load
Naver's HyperCLOVA X, which underpins search summarization, customer service chatbots, and a growing set of enterprise tools sold to Korean banks and insurers, sits closer to the center of the new rules because several of its enterprise deployments touch credit decisions directly. The company has set up a dedicated compliance unit and told partner banks it would supply the documentation needed for their own high-impact filings, a task that adds cost and lead time to deals that previously closed in weeks.
Kakao, whose messaging and payment platforms reach the majority of South Korean smartphone users, is registering its AI-driven credit scoring feature inside KakaoPay and a hiring-recommendation tool used by KakaoTalk's job-listing service. Both companies have said publicly that they support the law's intent while pushing, through the Korea AI Association industry group, for clearer technical guidelines on how bias testing results should be documented and shared with regulators.
A regional test case
South Korea's approach sits between the European Union's AI Act, which uses a similar risk-tiered structure but with a longer multi-year phase-in, and the lighter-touch guidance favored so far in Japan and Taiwan. Analysts tracking the rollout note that Korea's shorter grace period and concrete fine structure make it a closer real-world test of enforcement than the EU's still-unfolding timeline. Regulators in Japan's Ministry of Internal Affairs and Communications have said they are watching the Korean rollout as they finalize their own AI governance framework, expected to be presented to the Diet in 2027.