Korea AI Basic Act

Applies To

Overview

The Act on the Development of Artificial Intelligence and Establishment of Trust Basis (commonly referred to as the Korea AI Basic Act) was promulgated on 21 January 2025, following passage by the National Assembly in December 2024. It enters into force one year after promulgation. The Act establishes South Korea's first comprehensive legislative framework for artificial intelligence, creating a tiered regulatory structure anchored in risk classification. The legislation tasks the Ministry of Science and ICT (MSIT, the ministry covering information and communications technology) with primary oversight authority and creates a National AI Committee, renamed the National AI Strategy Committee by a January 2026 amendment. It also lets the Minister run an AI Safety Institute. The Act distinguishes between general AI systems and 'high-impact AI', defined as AI used in domains such as employment, education, finance, healthcare, criminal justice, and critical infrastructure, subjecting the latter to transparency, risk management, explanation, user protection, and human oversight duties. There is no pre-market conformity assessment. The Act also addresses generative AI, requiring watermarking (a marker showing content was made by AI) or disclosure where AI-generated content could cause public confusion. Developers and deployers of high-impact AI must conduct risk management activities, maintain documentation, and notify affected parties of significant AI-driven decisions. The Act includes provisions for government support of AI innovation, international cooperation, and the promotion of trustworthy AI standards.

Key Requirements

• •Classification of AI systems as general or high-impact based on use-case domain and potential for harm
• •Mandatory risk management frameworks for high-impact AI, including risk assessment, mitigation measures, and ongoing monitoring
• •Transparency obligations: users must be informed when interacting with high-impact AI systems
• •Human oversight requirements for high-impact AI decisions affecting individuals
• •AI-generated content disclosure or watermarking to prevent public deception
• •Documentation and record-keeping obligations for high-impact AI developers and deployers
• •Notification to affected individuals of consequential AI-driven decisions
• •A National AI Strategy Committee, and an AI Safety Institute the Minister may operate
• •Operators must review in advance whether their AI is high-impact, and may ask MSIT to confirm
• •Administrative fines of up to KRW 30 million for failing to notify users, appoint a domestic representative, or follow corrective orders

What Your Organization Must Do

• →Review in advance whether your AI falls into high-impact domains such as employment, finance, or healthcare.
• →Ask the Ministry of Science and ICT to confirm high-impact status where classification is unclear.
• →Build risk assessment, mitigation, monitoring and documentation records for any high-impact AI you develop or deploy.
• →Appoint a domestic representative in Korea if your company sells into the Korean market from abroad.
• →Add watermarking or disclosure to generative AI outputs that could confuse the public.
• →Allocate developer and deployer duties in contracts, and plan user notice for consequential AI decisions.

Playbook Guidance

Step-by-step implementation guidance for compliance teams.

Frequently Asked Questions

When does the Korea AI Basic Act take effect and what is the compliance deadline?

The Act was promulgated on 21 January 2025 and took effect on 22 January 2026. MSIT has indicated a one-year grace period before fines are imposed.

Which AI systems qualify as high-impact AI under the Korea AI Basic Act?

High-impact AI is defined by use, in areas such as employment, education, finance, healthcare, criminal justice, and critical infrastructure. It brings risk management, explanation, user protection, and human oversight duties.

Does the Korea AI Basic Act apply to foreign companies that do not have a Korean subsidiary?

Yes. The Act applies to developers and deployers of AI systems operating in or targeting the Korean market, regardless of where the company is incorporated. Multinationals with Korean-market products or services should assess extraterritorial exposure and allocate compliance responsibilities between developers and deployers in their contracts.

What conformity assessment process is required before deploying high-impact AI in Korea?

None is required: operators must check in advance whether their AI is high-impact and may ask MSIT to confirm. Impact assessments are an effort-based duty, and certification is voluntary.

How does the Korea AI Basic Act treat generative AI compared to the EU AI Act?

Both regimes impose disclosure obligations on generative AI outputs, but the Korea AI Basic Act focuses specifically on watermarking or disclosure where AI-generated content could cause public confusion, rather than the EU AI Act's broader transparency and copyright provenance requirements. Korean compliance programs should verify that watermarking or disclosure mechanisms are technically feasible and deployed before go-live.

What are the penalties for non-compliance with the Korea AI Basic Act?

Administrative fines of up to KRW 30 million, about US$21,000. They apply only to failing to notify users, failing to appoint a domestic representative, or ignoring corrective orders, and MSIT has indicated a one-year grace period.