LG AI Research has unveiled technology that automatically analyzes copyright and personal data infringement risks in AI training data, cutting legal verification costs to roughly one-thousandth of previous levels. The company said it won the National Intellectual Property Commission Chairman's Award in the copyright technology awards at the 2026 International Copyright Technology Conference (ICOTEC), hosted by South Korea's Ministry of Culture, Sports and Tourism, for its self-developed AI compliance agent EXAONE Nexus. The agent independently identifies copyright, licensing and personal data risks in large-scale datasets, automating screening previously done manually. It analyzes a single dataset 45 times faster than legal experts with 26% higher accuracy, at about 1/1,000 the cost of conventional methods. Its assessment uses 18 criteria derived from actual legal disputes and case law, scoring each from 1 point for high risk to 5 points for low risk and classifying datasets into seven levels from "very safe" to "very risky." The compliance framework reflects South Korea's AI Basic Act and the European Union's AI Act, with legal validity verified in cooperation with about 40 lawyers from major law firms in South Korea and abroad, including the United States. Risk assessment results for some 20,000 open datasets were released to academia and the research community. EXAONE Nexus is now a mandatory verification step for all AI models developed by LG AI Research, checking risks across the entire cycle from data collection and refinement to model development and service deployment.