Research on AI Policy and Issues in Key Areas in South Korea: Public Sector, Law Enforcement, Education and Social Welfare

Research on AI Policy and Issues in Key Areas in South Korea: Public Sector, Law Enforcement, Education and Social Welfare is the civil-society sectoral analysis published on 28 February 2025 by the Korean Progressive Network Jinbonet and the Institute for Digital Rights, with support from the Association for Progressive Communications (APC). The report maps the deployment of AI systems across South Korean public administration, law enforcement, education, and social welfare, analyses the regulatory standards and governance gaps in each sector, and situates the findings against the AI Framework Act passed by the National Assembly on 26 December 2024 — the legislation the Korean government framed as the country's foundational AI governance law, and that civil society assessed as systematically biased toward industry growth at the expense of the rights of citizens already affected by deployed AI systems.

Four sectors

Public sector

In the public sector the report finds AI chatbots widely deployed to assist civil servants with administrative tasks, but no comprehensive management system tracking which AI systems are in use, for what purposes, or under what governance standards. The absence of a systematic AI registry means that neither the public nor oversight bodies can determine the scope of AI deployment in public administration — the opacity problem that makes meaningful accountability structurally unavailable rather than merely absent in practice.

Law enforcement

The law-enforcement chapter documents AI-powered surveillance systems already in operation without adequate safeguards: the National Police Agency's real-time crime-prediction technologies; intelligent CCTV systems capable of facial recognition, behaviour recognition, and tracking across public space; and the Ministry of Justice's Immigration AI project, trained on 170 million biometric data points drawn from the South Korean population. The biometric data supply for these systems has its own documented history: in the pre-Act period the government had provided 170 million citizen records to private companies without individual consent, and the report situates the police AI deployment as the state's own parallel use of the same population-scale biometric infrastructure. The report's framing of the surveillance concern is that the intelligent CCTV system had already been developed for identification, tracking, and crime and behaviour prediction with no corresponding legal safeguards against its use against citizens.

Education

The Ministry of Education's AI Digital Textbook initiative — rolled out from 2025 — is criticised in the report for collecting excessive personal student data including after-school activities, lacking adequate data-protection safeguards, and having been developed and mandated without meaningful consultation with teachers or civil society. The report positions the digital-textbook rollout as a case study of government-led AI deployment proceeding ahead of governance design: the technology was mandated at scale before impact-assessment frameworks, consent mechanisms, or educator-participation processes were in place.

Social welfare

In social welfare, AI systems are used to monitor elderly health, detect fraudulent benefit claims, and manage welfare allocations — contexts where the affected population is among the most vulnerable and the stakes of algorithmic error are highest. The report finds that these systems lack consent protocols and fail to account for the needs of the populations they govern, raising concerns about both the adequacy of existing legal safeguards and the broader governance gap for AI making consequential decisions about access to public services.

AI Framework Act critique

The report's critical frame is the December 2024 AI Framework Act, which Jinbonet president Byoung-il Oh analyses as structurally inadequate on the following grounds: the act specifies no prohibited AI practices; the scope of "high-impact AI" providers is defined narrowly with inadequate penalties for violations; the act defines "those affected by AI" but contains no provisions for their rights or remedies; there are no disclosure requirements for general-purpose AI operators regarding training data; and an explicit security exemption removes AI used for defence or national security from the law's application entirely. The act passed with the unanimous support of both ruling and opposition parties, with only civil-society organisations on record as critical of its final form. Oh's framing is that in following an "allow first, regulate later" approach the act prioritises Korean AI industry growth over the protection of citizens whose lives AI systems are already reshaping — the same critique Korean civil society had been building since at least the 120-organisation civil society manifesto of May 2021 demanding a national AI supervision system, transparency and participation, AI risk assessment and control, and rights redress.

Recommendations

The report's recommendations address governance gaps sector by sector: mandatory AI registration establishing a public registry of AI systems deployed in public-sector contexts; mandatory human rights impact assessments prior to AI deployment; binding guidelines replacing voluntary frameworks; structured civil society involvement in AI policy processes; and strengthened accountability mechanisms for AI use in law enforcement. The companion Controversial Cases on AI in Republic of Korea (Jinbonet, January 2024) had already assembled the pre-Act catalogue of problematic deployments; the 2025 report converts that evidence base into a sectoral governance analysis and a critique of the legislative response.

Posture within the corpus

Research on AI Policy and Issues in Key Areas in South Korea is the corpus's first publication anchored in East Asia, filling the geographic gap the Association for Progressive Communications (APC)'s regional-research programme has been addressing systematically: the publications slate carries Southeast Asia in EngageMedia's Governance of Artificial Intelligence (AI) in Southeast Asia, sub-Saharan Africa in Legal Analysis of Laws, Policies and Government Strategies Relating to Artificial Intelligence in Kenya, Mauritius, Rwanda, South Sudan, Tanzania, Uganda and Zambia, Francophone West Africa in AfricTivistes's AI in Elections and the Challenge to Information Integrity, and MENA in SMEX's AI Investments in the Gulf, but had no East Asian anchor before this entry.

The South Korean entry is also significant for the corpus's legislative-moment register: the December 2024 AI Framework Act makes South Korea the second country globally to enact major AI legislation after the EU AI Act, and the Jinbonet-APC report is civil society's first systematic evidence base on how the pre-Act deployment landscape looked and where the Act's passage fell short. The publication's structural shape — a domestic civil-society organisation (Jinbonet, founded 1997, APC member) and a specialist research partner (Institute for Digital Rights) producing a multi-sector analysis for distribution through the APC global network — is consistent with the APC-anchored Global South regional-research pattern seen in the Southeast Asia and East Africa entries. Open Net Korea, the other APC-member digital-rights organisation in the corpus with a South Korean focus, has worked overlapping terrain on internet-rights litigation, strategic litigation, and AI-in-education advocacy; the Jinbonet report is the first publication-class research entry anchored on the Korean civil-society AI governance landscape as a systematic multi-sector analysis.