This article investigates how to ensure the quality of English translations of Korean court decisions within Korea’s institutional legal translation system. Although these translations have no authentic legal force in a monolingual jurisdiction, they constitute a major, regularly funded public project and present particular challenges. Translating court decisions between common law and civil law systems is among the most demanding forms of legal translation, making a systematic, standardized quality assurance framework essential. Yet institutions such as the Supreme Court Library of Korea lack explicit quality safeguards, raising the risk that low-quality AI output will enter public databases and, if reused as training data, entrench long-term quality defects. Against this backdrop, the article synthesizes key premises, quality indicators, and constraints for judgment translation discussed in Western legal translation scholarship and uses this framework to critically examine quality issues in current Korean practice, with particular attention to AI-generated translations of lower court judgments produced since January 2025. It then introduces the Court of Justice of the European Union (CJEU)—the world’s largest institutional provider of judicial decision translation—as a reference model. Drawing on CJEU annual reports, its AI strategy, tender specifications, and framework contracts, the study describes its internal translation and revision procedures, management of outsourced work, and use of terminological resources, translation memories, and AI tools. On this basis, the article proposes measures to improve institutional translation workflows in Korea, including coordinated development of style guides, termbases, and translation memories, alongside a more integrated approach to quality management for outsourced legal translation.