LG AI Research: Systematic Organization of AI Risks Through AI Ethics Seminar

Behind the infinite possibilities of AI technology lie various risks. To prepare for practical risks without exaggerating or downplaying them, accurate experience and knowledge of AI risk is essential.

LG AI Research operates an AI Ethics Seminar in which various experts, including AI researchers, business developers, legal professionals, and UI/UX designers, participate, analyzing the latest research trends to consider ethical aspects across the entire AI lifecycle.

Currently, the following tools and systems are being utilized to manage AI risks.

• AI Incident Database (AIID): a channel for voluntarily recording AI-related incidents
• AI Incidents Monitor (AIM): a real-time AI incident detection tool developed by the OECD

According to an OECD report, AI-related incidents surged 139% year-over-year in 2024, showing a rapidly increasing trend. Recent incidents have appeared across various areas, including privacy protection (111 cases), information security (148 cases), malfunction (91 cases), and human rights (86 cases).

Recently, there have been active attempts to classify risks in an integrated manner based on research such as the NIST AI Risk Framework, EPIC, Ferrara, and Weidinger. In particular, MIT has built the AI Risk Repository, containing 777 AI risks, leading the way in defining and systematizing risks.