Science ministry launches 80-member public AI safety monitoring corps
South Korea's science ministry has commissioned 80 citizens to report errors and risks in generative AI for a year, building on a month-long 2025 campaign that produced 464 valid error cases.

The science ministry said Thursday it had launched a public monitoring corps of 80 citizens tasked with finding and reporting errors and risks in generative artificial intelligence, expanding a one-off public campaign into a year-round reporting system.
The Ministry of Science and ICT and the AI Safety Institute commissioned the members, called "AI safety guardians," at a ceremony at the Glad Hotel in Yeouido, Seoul. The group will operate from September 2026 through August 2027.
The 80 were chosen from 138 applicants in two open recruitment rounds. They range from high school students in their teens to people in their 60s and older, and include office workers, homemakers, freelancers and students.
Their work covers four areas of generative AI failure: distorted facts, or hallucination; bias and discrimination; security vulnerabilities; and technical defects such as no response, answers in the wrong format or repeated replies.
The corps grew out of a campaign run for about 30 days from December 2025, in which the ministry and the institute asked the public to flag AI errors. It drew 625 submissions, of which 464 were confirmed as valid error cases.
That response convinced the ministry and the institute that a standing channel was needed rather than a temporary exercise.
Submitted cases will be verified for accuracy, risk level, reproducibility and usability before being confirmed. Confirmed cases will feed into AI safety evaluations and risk research and will be shared with the companies concerned.
The launch event drew about 50 people, including Um Yeol, director of the ministry's AI policy office, and Kim Myung-joo, head of the AI Safety Institute. It featured the awarding of commissioning letters, a briefing on operating plans and remarks from a representative guardian, followed by training on how to identify error cases and the standards for reporting them.
"Now that AI has become part of our daily lives for everyone, young and old, the environment for using it safely is something we all have to build together," Um said. "Each error our AI safety guardians find at the public's eye level will add up to become valuable groundwork for safer AI."
The launch comes as AI safety and trust have become a formal government task, with the AI Framework Act taking effect on Jan. 22, 2026.
What this article is based on
Every fact in this article can be checked against the primary documents below.
- Government과학기술정보통신부 과학기술 보도자료· Ministry of Science and ICT· accessed Nov. 14, 2026
- Government활동 기간· korea.kr· accessed Nov. 14, 2026
- Government정책 배경(업황 맥락)· korea.kr· accessed Nov. 14, 2026
- Government참여 구성· korea.kr· accessed Nov. 14, 2026
© New Era Korea Daily. All rights reserved.