Domestic artificial intelligence (AI) models are spreading across industrial sites, from financial investment analysis to robot intelligence and judicial administration.

The Ministry of Science and ICT said on Sunday that LG AI Research is working with Kiwoom Securities to build a finance-focused AI agent. Based on Exaone BI (Business Intelligence), it is implementing an "explainable AI investment" service that predicts stock investment scores and explains the basis for its judgments. It aims to raise reliability by combining it with financial data from the London Stock Exchange Group (LSEG).

Lee Hwa-young (이화영), an executive director at LG AI Research, described it as "a structure in which multiple expert agents collaborate organically to carry out the entire process from data analysis to prediction and report generation."

Motif Technologies is also targeting the financial sector. It will develop a finance-specialised large language model (LLM) by combining financial data and structured knowledge from DeepSearch with its own AI model. It plans to develop it into agentic AI that directly assists specialised tasks such as corporate analysis, asset management portfolio optimisation and evaluation of acquisition targets in mergers and acquisitions. It will supply it in an on-premises environment to about 50 financial companies linked to DeepSearch.

Upstage is teaming up with physical AI startup Realworld to pursue robot intelligence. It is pushing work to link "Solar Open VLM", a proprietary AI model to be developed with government support, with Realworld's robot foundation model (RFM) "RLDX-1". The goal is to expand into physical AI sites where robots understand real situations and execute high-degree-of-freedom movements.

SKT is working with domestic AI startup Liner to focus on improving the reliability of AI-based search. The key is to reduce hallucinations by combining Liner's high-precision retrieval-augmented generation (RAG) technology with its proprietary AI model "A.Dot X (A.X) K1". Liner is responding by designing a real-use-based dataset, a ranker-model-based evaluation system and a sentence-level verification module.

Naver Cloud has entered agricultural AI. Together with Daedong AI Lab, it will build an "agriculture-specialised AI agent" trained on domestic soil, climate and crop data. It aims to provide AI-based services from writing farming logs and tailored cultivation management to agricultural consulting and distribution and operations management. It plans to pursue automation across the entire agricultural process by combining Naver Cloud's full-stack AI technology with Daedong's agricultural AI and robotics technology.

KT is working on a project with the Supreme Court to build an AI system to support trials. It is designed to provide answers based on vast legal information such as laws, precedents and decision cases, and to allow users to check the underlying laws and precedents as well. KT plans to apply its own AI model, "Mideum 2.0", to the system.