[단독] 정부 첫 ‘고영향 AI’ 판정은 ‘범죄 수사용 얼굴인식’ 기술

The Ministry of Science and ICT has officially designated two artificial intelligence technologies as "high-impact AI" for use in criminal investigations. This marks the first such case since the implementation of the Artificial Intelligence Basic Act last January. The term "high-impact AI" refers to artificial intelligence systems used in 10 different sectors, including healthcare, crime investigations, and public services, that could significantly impact a person's life, bodily safety, or fundamental rights.

According to a report submitted by the Ministry of Government Legislation on May 15th, the Ministry of Government Legislation processed 16 requests for "high-impact AI verification" between January 22nd and March 1st, 2023. Out of these, 2 were deemed to fall under the high-impact category, 12 were deemed not to, and 2 were rejected.

Businesses must review their services to determine if they fall under the high-impact AI classification, and if uncertain, can request verification from the Ministry of Government Legislation. If classified as high-impact AI, service providers must disclose to users that they are using AI-based services, establish risk management and user protection measures, and adhere to supervision plans for managing individuals.

The Ministry of Government Legislation plans to investigate the implementation status of businesses through industrial surveys starting this month. 이주희, a member of the opposition party, emphasized the importance of clear standards for the high-impact AI classification, stating that the issue directly relates to citizens' lives, safety, and fundamental rights.

The Ministry of Government Legislation has pledged to ensure the system is firmly established during the minimum 1-year probationary period following the implementation of the law. The two cases deemed as high-impact AI both involve facial recognition technology for criminal investigations. The 3D facial recognition technology used in one case, which repeatedly identifies suspects, received the high-impact classification due to its analysis and utilization of biometric information.

Additionally, one AI system combining face recognition, facial prediction, deepfake detection, and facial restoration was partially classified as high-impact AI. The facial recognition technology was deemed high-impact because of the potential for violating fundamental rights through biometric analysis. However, AI technology used in processing resident registration card information during the system process was deemed not high-impact, as it was considered a simple automation of routine business.