The pace at which artificial intelligence is advancing is astounding―so rapid that governance struggles to keep up. Social risks loom in the gap between technological progress and regulatory response. While AI is rightly celebrated for its transformative potential, we must also heed warnings about the dangers it poses. One such warning comes from Geoffrey Hinton, a University of Toronto professor and 2024 Nobel laureate in physics, who told the BBC last December that there is a 10 to 20 percent chance AI could wipe out humanity within 30 years. Social security is one of the public service areas where AI technologies are most widely applied. In Korea, AI has been used to identify at-risk households and support job placement. Albeit slowly, regulations have been put in place to both harness AI’s immense potential and mitigate its risks. A key example is the Basic Act on the Development of Artificial Intelligence and Establishment of Trust (the Basic Act on AI), passed in December 2024. Yet, it is hardly the case that social security received sufficient attention in the course of drafting this law. According to one index, Korea currently ranks as the world’s sixth most advanced AI powerhouse. But policy conversations around regulation, particularly concerning AI use in social security and related fields, are progressing at a glacial pace. This month’s issue of Health and Welfare Forum spotlights ‘Artificial Intelligence and Social Security,’ with four feature articles devoted to the theme. The first reviews the use of AI in Korea’s social security system. The second examines applications of AI in US social security, with a focus on Executive Order 14110, an exemplar AI regulation from the Biden Administration. The third provides a detailed account of Korea’s Basic Act on AI, assessing both its significance and its limitations. The final article explores the benefits and risks of applying AI in social security and offers policy implications.

This article examines current applications of ICT in Korea’s social security programs and discusses actual cases of AI application in their administration. ICT is used widely across the social security system in both administration and service delivery. The AI-assisted job matching service, implemented to reduce search costs for job seekers and lessen the likelihood of occupational mismatch, is on track for expansion and refinement with government support. The existing system for identifying social security gaps has advanced by incorporating a wider range of linked datasets and enhanced data-analysis algorithms, especially after the tragic ‘mother-and-two-daughters’ incident in 2014 in Songpa-gu, Seoul. While AI applications in social security programs can bring benefits such as efficiency, cost reduction, and personalization in service delivery, they also pose risks, including added workloads, data-protection issues, and accuracy concerns―all of which require policy preparedness.

AI technologies are widely used in social security systems worldwide. In many welfare states, including the US, the UK, Austria, and New Zealand, AI-powered tools―such as chatbots, machine learning-based models for identifying eligible individuals, and automated claims systems―assist in various aspects of health and social welfare policy administration. By contrast, initiatives like the Dutch SyRI, Australia’s Robodebt, and Denmark’s Gladsaxe have raised concerns regarding discrimination, human rights violations, and data leaks. Executive Order 14110, issued during the Biden Administration, exemplifies regulatory responses to such issues. Although later rescinded following the advent of the second Trump Administration, this order, featuring eight guiding principles for safe and equitable AI development, along with concrete action directives for the Department of Health and Human Services and other key federal agencies, still resonates as an international standard in AI regulation. As for social security in Korea, it is necessary to establish AI regulations taking into account the issue of explainability, the right to appeal, and the need to improve the algorithm audit system. It is also important to strengthen organizations and ensure sufficient staffing for verifying and improving the quality of social security data, as well as managing associated risks.

On January 21, 2025, Korea enacted its Basic Act on AI, the second of its kind in the world. Korea’s Basic Act, while combining the two types of AI-related legislation―the European model, which focuses on protecting basic rights and the US model, which prioritizes industrial promotion―leans slightly more toward industrial promotion. It specifies national agencies and policy support for the development of AI-related industries and sets out general provisions, such as definitions, but offers little guidance on specific areas where AI applications are active or imminent, such as social security. In contrast, the EU’s AI Act and the US Executive Order on AI include provisions on social security, though insufficiently, leaving room for developing policies specific to AI use in this area. Korea’s Basic Act on AI delegates policymaking details to subordinate legislation, under which the National AI Commission, established by the law, is tasked with bringing together stakeholders representing the various segments of the social security sector.

In this article, I provide a comprehensive overview of the benefits and risks of AI applications in the area of social security and discuss their policy implications. AI has been applied across various dimensions of social security administration in Korea and elsewhere―notably in identity verification, eligibility screening, benefit calculation, fraud detection, risk classification, personalized services, case management, administrative process optimization, and policy evaluation. Using AI in these areas has proved beneficial in increasing administrative efficiency, improving timeliness and accuracy of benefit payments, reducing unmet needs, tailoring services to individual needs, and establishing databases for policy evaluation. However, AI applications also carry risks involving personal information breaches, data bias, the opacity of algorithmic decision-making, potential commercial exploitation of information, and the problem of explainability. The implications drawn point to the need to improve data quality, promote standardization, establish integrated governance, revise legal frameworks to include provisions on social security, create specialized organizations, and monitor international trends.

In this article, I analyze the dynamics of social withdrawal among young Korean adults and their needs, drawing on micro-data from the Youth Survey for 2022 and 2024, and present policy suggestions for more effective youth support. The share of socially withdrawn individuals, those who, in near-complete detachment from social engagements, rarely leave their homes, rose from 2.4 percent in 2022 to 5.2 percent in 2024. This increase does not necessarily mean that more young adults have become reclusive; rather, it suggests that some who had previously remained hidden in isolation are, in a sense, stepping forward, voicing their struggles, and seeking support for recovery and independent living. I analyze the living conditions and needs of these socially isolated young adults, a newly recognized vulnerable group, and propose the following policy measures: nationwide expansion of Youth Future Centers and the systematization of support services; tailored assistance to promote recovery and socioeconomic independence; regional expansion of initiatives through hub institutions; training and improved treatment of specialized personnel; the cultivation of public awareness; and the establishment of a medium- to long-term policy roadmap.