LG Conference Shares 15 AI Safety Cases

AI Times KoreaFriday, November 6, 2026
LG Conference Shares 15 AI Safety Cases
  • •LG’s safety and environment conference shared 21 innovations, including 15 AI use cases.
  • •NC AI will run a VARCO experience cafe during G-Star 2026, while DeepBrain AI tested financial automation in Vietnam.
  • •SelectStar’s FinRed improved expert-judgment agreement by 11.54 percentage points and cut unsafe answers misclassified as safe by about 57%.

LG said it held its group safety, health and environment conference, LG SHEC, at LG Sciencepark in Magok, Seoul. More than 350 safety and environment staff from its affiliates attended and shared 21 innovations, 15 of them involving AI. LG Electronics presented an integrated monitoring system that directly controls equipment when AI detects danger. LG Chem introduced an AI agent that reviews process safety management documents.

NC AI said it will operate a cafe near BEXCO in Busan during G-Star 2026. Visitors can try VARCO 3D, which creates 3D models from images or text, and VARCO Sound, which converts voices into creature sounds. Korea Deep Learning said it conducted a proof of concept with Shinhan Vietnam Finance on financial task automation through Shinhan Futures Lab Vietnam, Shinhan Financial Group’s startup development platform. The project used Deep Agent to extract and structure document information and connect it to work processes. The company plans to expand into overseas financial markets, including Vietnam.

Grepp won awards in both the assessment and edtech categories at the A-ATP 2026 Awards, hosted by the Asia Association of Test Publishers. It also introduced AI assessment and proctoring services planned for release next year. Grepp is developing an agentic AI assessment to measure software development skills using AI, as well as an AI exam proctoring camera equipped with its own AI model, with commercialization targeted for next year.

Selectstar said a paper it co-authored with the Financial Security Institute and KAIST was selected for an oral presentation at IEEE ICDM 2026. The paper proposes FinRed, a framework for assessing the safety of large language models specialized in finance. Compared with existing evaluation methods, FinRed raised agreement with financial security experts’ judgments by 11.54 percentage points and reduced by about 57% cases where dangerous answers were incorrectly judged safe.