LG AI Research 534

Alongside the rapid advancement of AI models, reliability and safety issues in the data training process are emerging as key challenges. In particular, modern AI datasets have a hierarchical structure in which numerous sub-datasets are intricately intertwined, making it difficult for humans to directly track the origins and rights relationships of the data.

LG AI Research has developed Agent AI technology to address this problem by tracking the lifecycle of datasets and comprehensively analyzing legal risks. This technology automatically detects changes in rights relationships that occur as data is redistributed and combined.

The newly released NEXUS is a platform that lets users directly explore this Agent AI technology. Its key features are as follows.

• Hierarchical Structure Analysis: Recognizes a dataset not as a simple collection of information but as a complex tree structure, automatically analyzing how the legal risk of each node affects its parent nodes.
• Risk Detection: Detects license conflicts, inconsistencies in rights relationships, and the presence of personal information step by step, providing detailed risk assessments.
• Based on EXAONE 3.5: The data compliance solution applied to NEXUS is based on the EXAONE 3.5 model and consists of three core modules—Navigation, QA, and Scoring—to perform accurate and efficient analysis.