South Korea - AI Technology Framework (RAI-KR-NA-BATRMXX-2020)

Overview

The Basic Act on AI Technology (인공지능 기술 기본법안), proposed by Representative Min Hyung-bae on 2020-10-29 (recorded in the National Assembly as 의안번호 2104772), was a foundational initiative intended to provide a statutory framework for South Korea's AI policy: national planning, industry promotion, institutional governance, ethical safeguards and trust-building measures. The draft focused on establishing a national AI planning cycle, a central governance committee and a national AI centre for technical support and cluster development. Although the bill was debated in the Science, ICT, Broadcasting & Communications Committee, the proposal lapsed at the end of the 21st National Assembly and did not become law. For contemporaneous references and later legislative consolidation that drew on prior bills, see analysis by the National Assembly research/policy archives and legislative tracking resources such as KCI / National Assembly research publications and catalogues of AI bills used during committee consolidation processes.

Definitions

The draft defined core terms to set scope and obligations. Typical definitions included "artificial intelligence" (broadly understood to include machine learning, deep learning, natural language processing and related algorithmic systems), "AI developer/provider" (entities engaged in development, deployment, distribution or commercial offering of AI products or services), "high-impact AI" (systems whose functioning can affect life, health, basic rights or essential economic outcomes), and "AI industrial cluster" (designated geographical or functional concentration of AI firms, research institutions and support organisations). The draft sought to align terminology with existing Korean statutes (e.g., 지능정보화/지능정보화기본법 and 개인정보보호법) while creating a tailored lexicon for governance and promotion.

Governance and Institutional Framework

The bill proposed a tripartite governance architecture: (1) a National AI Committee (or similar high-level coordinating body) established by statute to review and approve national plans and supervise cross-ministerial policy, (2) a National AI Centre or technical secretariat to provide R&D coordination, standards support, testing capacity and training programs, and (3) ministerial responsibilities (principally the Ministry of Science and ICT — 과학기술정보통신부) to implement policy, issue guidance and liaise with industry and local governments. The bill delegated detailed procedural and organizational design to subordinate rules but mandated regular basic-plan cycles (e.g., every three years) and explicit budgeting assignments. See the legislative consolidation materials and committee reports that built on early bills: Lawmaking - AI basic law (consolidated) for later institutional choices informed by 2020–2024 drafts.

Key Focus Areas

Core policy strands in the 2020 draft included: (a) Promotion & Industrial Policy: statutory powers for state-level support of AI R&D, workforce training, subsidies/grants, and designation/support of AI industrial clusters to foster scale and regional economic development; (b) Safety & Trust: ethical guidelines embedded into statutory obligations, privacy safeguards, and risk-management measures for systems that may impact safety or rights; (c) Transparency & Accountability: requirements for documentation, explainability objectives for developers and voluntary/mandatory impact assessments for high-impact AI; (d) Standards & Certification: support for testing facilities, standards development and private/ public certification to enable conformity assessment; (e) Data & Privacy: coordination with South Korea's 개인정보보호법 and measures to promote privacy-preserving data practices and secure data infrastructure; (f) International engagement: provisions anticipating the need to align with global AI governance developments. The bill intentionally combined promotion and safeguards — a theme that recurred in subsequent consolidated legislation across 2022–2024.

Implementation Framework

The draft envisaged a phased implementation model: national basic plans at multi-year intervals, a governance committee to approve strategies and budgets, and an operational centre to manage technical programs. Implementation tools named in the draft included targeted R&D funding programs, incentives for private-public partnerships, designation procedures for AI clusters, support for standardisation and testing, and guidance for impact assessment modalities. The bill relied on subordinate regulations (enabling decrees and ministerial ordinances) to specify thresholds, procedural steps for designation/certification, and enforcement modalities — a common legislative approach to balance a high-level framework with flexibility in technical detail. For how these design choices played out in later consolidated law, consult the lawmaking/consolidation materials and committee '검토보고' that summarise how early drafts such as 의안번호 2104772 were integrated.

Monitoring and Evaluation

The draft required periodic reporting and review: the National AI Committee would be mandated to review national plans and publish progress reports; the National AI Centre (or equivalent) would collect programme-level metrics (R&D outputs, workforce indicators, cluster performance) and support independent evaluation. For high-impact AI systems, the bill anticipated post-market monitoring obligations and encouraged transparent publication of evaluation results and corrective measures. These monitoring duties were designed to be coordinated with sectoral regulators (e.g., financial regulators, health regulators) to avoid gaps or conflicting oversight.

Penalties, Liability, and Appeals

The 2020 draft foregrounded obligations and administrative oversight rather than novel criminal sanctions; it envisaged administrative measures (orders to remedy, suspension of provision) and civil liability mechanisms where AI systems caused harm, with delegated authority for fines or administrative penalties to be specified in subordinate rules. The draft also emphasised procedural protections and remedies for affected persons (complaint mechanisms, rights to explanation or redress). The bill left much of the detailed liability architecture and penalty levels to implementing regulations, consistent with its status as a "basic" act. Later consolidated legislation clarified and expanded some enforcement mechanisms drawing on earlier drafts' aims.

Relationship to Other Instruments

The draft was explicitly designed to operate alongside and connect to existing legal frameworks — most notably the 지능정보화 기본법 (Intelligent Informatization Basic Act), the 개인정보 보호법 (Personal Information Protection Act), and sectoral regulators' statutes (healthcare, finance, transport). It proposed coordination clauses to avoid regulatory overlap and to identify lead agencies for sectoral coordination. The draft also referenced the need for alignment with international standards and ongoing domestic standardisation and guidance activities. See committee-level review materials and consolidated law materials for cross-references.

International Alignment

Although principally domestic in scope, the bill signalled intent to align South Korea's governance approach with emerging international practices on AI safety, transparency and high-risk system oversight. The draft noted global legislative trends and the need for interoperability with international technical standards and conformity assessment regimes. Subsequent legislative consolidation in 2024–2025 explicitly referenced EU and US policy developments, illustrating the continuity of the policy trajectory from early drafts such as this 2020 proposal to later enacted measures. For comparative policy context consult research/analysis sources and the 2024 consolidated law materials.

Implementation Timeline

Sources and References