KT's Trial Support AI System applies AI technology across core trial-related tasks, including similar case searches, issue analysis, and document drafting. /Courtesy of KT

KT will unveil results from public artificial intelligence (AI) pilots at MWC 26, the world's largest mobile communications expo, in Barcelona, Spain.

KT said on the 2nd that it will introduce an AI platform that supports the Supreme Court's trial operations and Gyeonggi Provincial Office of Education's AI teaching-learning platform "Hi-Learning" at the event.

KT's "trial support AI system" applies AI technology across core trial-related tasks such as similar-case search, issue analysis, and document drafting, with the goal of boosting the productivity of judges, judicial researchers, and court officials. The company plans to gradually implement optimal databases and customized AI services based on its commercial know-how in LegalTech AI.

KT won the "AI platform for supporting trial operations" project from the Supreme Court last year. The project will proceed in four stages over a total of 45 months and is designed as a high-difficulty task that meets 128 requirements across 12 areas for real users. Currently, the "intelligent legal information search" area, corresponding to the first year, is in a pilot phase, and services will be expanded sequentially.

The trial support AI platform includes a range of services, such as intelligent semantic search and research, summary and issue analysis, review and analysis of new cases (newly filed cases), support for drafting judgments and review reports, and personal information extraction and de-identification. Among these, at MWC26 this year, KT will focus on the "intelligent legal information search service," showcasing contextual understanding of legal documents and advanced search and analysis capabilities.

The service combines vector databases with large-scale retrieval-augmented generation (RAG) technology to enable similar-case searches and precise document-based Q&A.

KT said it will apply an optimal lineup of large language models (LLMs) for trial support by using two models: "Believe:um K 2.0" and Meta's "Llama K." It pre-trained on about 20 billion items of high-quality legal data to secure its own language model quality, and based on this, recorded strong performance in evaluations of legal knowledge language models.

KT will also build databases based on more than about 8.38 million source documents—totaling 25 terabytes (TB)—including judgments, complaints, briefs, answers, transcripts, and literature, leveraging its understanding of legal documents and search expertise.

In addition, since 2023, KT has operated "Hi-Learning," an AI-based teaching-learning platform, with the Gyeonggi Provincial Office of Education, Korea's largest education office. Hi-Learning is an AI-based digital platform that teachers and students can use across the entire class process, supporting teachers' digital-based class operations and students' self-directed and personalized learning. It also offers a variety of content and external Edtech services.

Yoo Yong-gyu, head of the public projects division (senior vice president) at KT Enterprise, said, "KT will continue to expand services tailored to the practical needs of public institutions by leveraging various AI models such as 'Believe:um 2.0,' becoming a trusted AX (AI transformation) partner in the public sector."

※ This article has been translated by AI. Share your feedback here.