South Korea - AI Regulation Bill (RAI-KR-NA-ALGARIN-2021)

Overview

This bill, titled the "Act on Algorithms and Artificial Intelligence," was introduced by Representative Yoon Young‑chan and co‑sponsors on 24 November 2021 (Bill No. 2113509). Its stated purpose was to foster development and commercialization of algorithm and AI technologies while minimizing negative impacts by establishing legal definitions, governance structures, user protections, and industry support measures. The bill focused particularly on the concept of "high‑risk AI" and institutional oversight. The legislative record shows the bill advanced through committee stages, public hearings and subcommittee review between 2021 and early 2023 but was ultimately not enacted; it was recorded as lapsed (expired at the end of the legislative term) on 29 May 2024. Full text and bill information are available from the National Assembly e‑legislation / bill information system.

Definitions

The proposal included statutory definitions intended to delimit coverage and obligations. Core terms defined in the draft law included: "algorithm," "artificial intelligence," "high‑risk artificial intelligence" (described as high‑impact systems that can materially affect life, safety, or fundamental rights), and "AI development operator" (actors engaged in developing or providing AI systems). Those definitions would determine when governance, transparency, and protective measures applied—particularly for systems with significant societal impact.

Governance and Institutional Framework

The bill proposed a multi‑layered governance structure centered on state oversight and dispute resolution mechanisms. Key institutional elements included:

• A High‑Risk AI Deliberation Committee (or equivalent body) under the Prime Minister to deliberate and decide on matters relating to regulation of high‑risk AI, to set basic principles, and to guide related policies.
• An Algorithm & AI Dispute Resolution Committee to adjudicate disputes arising from algorithmic/AI services and outcomes.
• Provision for private autonomous AI ethics committees at research and development institutions and among developers to oversee compliance with development principles and ethical norms.
• Government duties to promote R&D and commercialization of algorithm and AI technologies, balanced with targeted oversight for designated high‑risk systems.

Key Focus Areas

• Definitions and scope: Clarifying what constitutes an "algorithm," "artificial intelligence," and especially "high‑risk AI" to identify systems subject to enhanced oversight.
• Risk‑based regulation: Targeted regulatory treatment and special obligations for AI systems designated as high‑risk because they substantially affect life, body, safety, or fundamental rights.
• User protections: Rights for users of high‑risk AI, including the right to request explanations, the right to object to outcomes, and the right to refuse use where appropriate.
• Transparency and accountability: Duties for AI development operators to enable explanation, provide information to users, and participate in dispute resolution processes.
• Liability and remedies: Civil liability for harm caused by high‑risk AI systems and administrative oversight mechanisms to provide recourse for affected individuals.
• Dispute resolution and oversight: Establishment of administrative committees for adjudication and deliberation rather than extensive criminal penalties in the draft text.
• Support for innovation: Government measures to promote R&D and support commercialization while maintaining safeguards for high‑risk applications.

Implementation Framework

Implementation under the draft law would have combined statutory duties for AI development operators with administrative oversight and government support measures. Operators of high‑risk AI systems would be required to implement procedures enabling explanation to users, to respect user objection and refusal rights, and to respond to disputes through the proposed dispute resolution committee. The government would be mandated to promote research and development and to facilitate commercialization of algorithms and AI more broadly, while the High‑Risk AI Deliberation Committee would set principles and oversee designation and management of high‑risk systems. Specific operational details, such as implementing rules, administrative procedures, standards for explanations, and penalty schedules, were to be addressed in subsequent implementing regulations or later legislative negotiation.

Monitoring and Evaluation

The draft emphasized administrative oversight and committee review as primary monitoring mechanisms. Monitoring and evaluation would have been carried out through the proposed High‑Risk AI Deliberation Committee and the Algorithm & AI Dispute Resolution Committee, which would review high‑risk designations, supervise compliance measures, and hear disputes. The bill record also shows committee hearings, public hearings and subcommittee review as part of the legislative monitoring process. Because the bill lapsed on expiry of the legislative term, no statutory monitoring regime from this bill entered into force; subsequent monitoring and evaluation efforts were pursued through later, separate legislative initiatives and consolidation efforts.

Penalties, Liability, and Appeals

The draft law prioritized civil remedies and administrative review over extensive criminal sanctions. Key features included provision for civil liability where harm was caused by high‑risk AI systems, and administrative oversight through the proposed deliberation and dispute resolution committees. The draft text indicated users could seek damages from AI development operators and access administrative dispute resolution mechanisms. Specific penalty levels, criminal sanctions (if any), and procedural appeal routes were not fully specified in the material provided and were expected to be addressed in implementing rules or subsequent legislative negotiation.

Relationship to Other Instruments

The bill explicitly referenced the broader intelligent information/AI legislative context (for example, the existing Intelligent Information / 지능정보화 legal framework) and was intended to function alongside existing sectoral laws governing privacy, safety, and consumer protection. It sought to provide a more specific statute focused on algorithms and AI, requiring coordination with data protection, consumer rights, and safety laws where obligations and remedies intersect. The legislative record also notes that subsequent consolidation and separate AI framework legislation (often referenced as an "AI Basic Act" or similar consolidated framework) was pursued in later National Assembly activity; this particular bill lapsed and therefore did not become part of the statutory framework.

International Alignment

The draft text as described in the legislative record did not set out explicit alignment with particular international instruments. Its focus was on establishing a Korean statutory approach to algorithm and AI governance and on coordination with existing national sectoral laws. Subsequent legislative efforts in Korea have considered international developments and comparative frameworks in the wider policy debate, but this bill itself is not recorded as establishing formal cross‑jurisdictional alignment mechanisms.

Implementation Timeline

Sources and References