Welfare Blind-Spot Discovery System — South Korea's AI Search for Citizens Who Qualify for Benefits But Never Applied

Following the 2014 death of a mother and two daughters in Seoul's Songpa district, who took their own lives after falling through gaps in Korea's welfare system, the government built a nationwide welfare blind-spot discovery system run by the Social Security Information Service under the Ministry of Health and Welfare, which cross-references data that citizens have already given the state rather than waiting for people to apply.

In June 2026 the ministry expanded the linked data to 47 categories, including financial data on rejected microfinance applicants and immigration records for children who may be abroad, and said the system now uses AI to learn complex crisis patterns and detect subtle signals that were difficult to capture through existing manual analysis, while since 2025 an AI phone service has made the first round of outreach calls to flagged households.

Official data submitted to the National Assembly show that 1.424 million people were identified as being in a welfare blind spot in 2024 alone, but 591,806 of them (41.6%) ended up receiving no welfare service at all, and only 17.7% of those helped received public benefits directly, with the rest referred to private charities, while the AI phone outreach connected with 63.9% of flagged households, of whom 65.7% went on to receive some form of support.

An MIS Quarterly case study on the system's core algorithm found that, despite the machine-learning model outperforming alternatives, many local governments still run parallel rule-based and manual screening because caseworkers distrust and do not understand the ML outputs, a governance gap the ministry has not yet closed even as it keeps adding new, more sensitive data sources.