With the judiciary piloting the opening of a trial-support artificial intelligence system (AI), judges can now search legal information and use it in trials by conversing with the Generative AI. The National Court Administration said it is providing the service through the judiciary's own platform, not an external Generative AI or large language model (LLM).
According to legal sources on the 18th, the National Court Administration built the judiciary's own AI platform and recently piloted the opening of the trial-support AI system. A court official said, "The key will be how quickly it can assess case records and how much insight it can provide to the panel."
The first-phase task unveiled this time is an intelligent legal information search and research system. A large language model refers to an AI model that learns vast amounts of text and generates sentences.
When this system is queried about legal knowledge or the issues in a particular case, it compiles and summarizes answers based on highly similar precedents, relevant statutes, and literature. A list of References is presented on the screen, and the original text can be checked through a view-original function.
The primary materials used are Supreme Court precedents and the full corpus of judgments from the past 13 years that were entered into the existing system. Statutes, Supreme Court rules, decisions, authoritative interpretations, practical summaries, and annotated commentaries are also used.
The National Court Administration plans to enable immediate retrieval and use of newly rendered judgments as soon as they are registered in the internal system. While there have been legal tech services centered on law firms, it cited as a differentiator that the system was built on internal infrastructure without relying on external models.
The court is simultaneously working to improve answer accuracy in the first-phase system and to advance its structure. It is also aiming, as early as within the year, to develop the second phase, which will analyze filed case records to organize the gist and issues.
The second phase will analyze filings such as complaints, preparatory briefs, and answers to distill the essentials. Once phases three and four are completed, functions to check for logical errors or awkward phrasing in draft judgments and to assist with verifying service addresses are also under review.
The National Court Administration presented a plan to use AI throughout the entire trial process, from filing a complaint to completing the judgment. As increasingly voluminous case records are cited as a cause of trial delays, the intent is that once the support system takes root, it could also help ease delays.
This year, the National Court Administration established AI policy and created one judicial artificial intelligence reviewer to examine intelligent information technologies such as AI and big data. In Apr. last year, it launched the Judiciary Artificial Intelligence Committee as an advisory body to the chief of the administration, continuing discussions on adopting AI.
At the end of last year, the committee presented a step-by-step roadmap for the next four years with the goal of realizing judicial justice through human-centered AI. In this year's New Year's address, Chief Justice Jo Hee-de announced a policy intent to enhance work efficiency by developing an AI-based trial-support model.