NC AI Implements 'Safeguard' Technology to Block Risks in Customer Service Chatbot

NCER Chatbot Equipped with Triple-Layered Security Technology
First Comprehensive AI Safety System Commercialized in Korea

On September 30, NC AI announced that it has officially implemented its proprietary artificial intelligence (AI) safety technology, 'Safeguard,' into NCSOFT's customer service chatbot, 'NCER (Answer).' This marks the first time a domestic AI company has introduced a comprehensive AI safety system into a commercial service.

NC AI Introduces 'Safeguard' to Customer Service Chatbot. Provided by NC AI

NC AI's Safeguard technology operates on a triple-layered security system composed of Red, Blue, and Purple teams. The Red Team researches new malicious attack patterns such as jailbreaking, while the Blue Team develops defense technologies against these threats. The Purple Team integrates the findings from both teams and reflects them in policy through a cyclical structure.

In particular, NC AI collaborated with NCSOFT's Publishing Coordination Center to establish a 'Chatbot Negative Regulation Policy' tailored to the characteristics of the gaming industry. This policy includes inappropriate content related to game services, loopholes involving paid items, and prohibited activities.

NCER is a service launched last year to enhance customer service (CS) efficiency and customer satisfaction using AI technology and game data. It handles real-time user inquiries, bug reports, and suggestions for major games.

With the integration of Safeguard, NCER blocks general inappropriate content such as discrimination, hate speech, profanity, and obscenity, as outlined by ML Commons and international benchmark studies, while also addressing the unique safety requirements of the gaming industry. The system detects and immediately blocks a wide range of potential risks that may arise in real-time chat environments. In practice, a spam filtering system is in operation for 'Lineage W' and 'Lineage 2M,' detecting and blocking advertising patterns in 13 languages.

NC AI plans to expand its text-based safety technology to multimodal content detection based on vision-language models (VLM). The goal is to build a next-generation AI safety system that comprehensively blocks inappropriate content in images and videos. In partnership with AIWorks, NC AI is also developing a safe AI response system for high-risk situations. The company expects these efforts to contribute to the establishment of national AI safety standards.

NC AI's technology is expanding beyond game chatbots into various industry sectors. In the new role-playing game (RPG) 'Aion 2,' the company introduced text-to-speech (TTS) technology capable of natural language-based animation generation and emotional expression.

Lee Yeonsoo, CEO of NC AI, stated, "The implementation of Safeguard technology is an important starting point for NC AI to secure technological leadership in AI safety and to solidify its position as a responsible AI development company. We will continue to fulfill our mission as a leading Korean AI company by achieving technological independence, industry validation, and global contribution simultaneously."

This content was produced with the assistance of AI translation services.