Utilization and Research Trends of Legal Large Language Models in Korea : Focusing on Hallucination and Security Risks

Focusing on hallucination and information security risks, this study examines the current status and research trends of largelanguagemodels(LLMs) specialized for law. LLMs have already been introduced into practice worldwide across various domains—including contract review, case law search, and document drafting—primarily by large law firms and global legal information companies, with reported improvements in efficiency and accuracy. However, despite the growth of the Korean legal tech industry, full-scale commercialization in Korea has been delayed due to factors such as limited data accessibility, regulatory barriers, and the conservative nature of the legal profession. We analyze key hallucination and security issues by synthesizing and organizing existing literature, and propose technical, institutional, and societal response directions. The adoption of cutting-edge techniques like Retrieval-Augmented Generation and fine-tuning, establishing standards and accountability frameworks for legal AI use, enhancing legal community awareness, and developing ethical guidelines are confirmed as essential in the areas of Korean legal tech. In summary, the true value of legal LLMs can only be realized when technological improvements, institutional enhancements, and expanded social acceptance are organically integrated.