Naver Cloud is participating in a government initiative to develop a security-focused artificial intelligence (AI) model with 700 billion parameters.
Naver Cloud announced on September 3 that its consortium has been selected as the contractor for the 'Specialized AI Foundation Model (Cybersecurity Sector)' project.
The consortium consists of 33 institutions from academia, industry, and government. It aims to build an AI model optimized for the domestic security environment and to validate and enhance it in key national facilities and major industrial sites, thereby boosting the competitiveness of domestic security AI.
Before officially starting the project, the consortium has already begun pre-training the security-focused AI model using 4,000 graphics processing units (GPUs). This includes 256 GPUs provided by the government, in addition to Naver Cloud's 4,000 B200 GPUs and 256 H200 GPUs from LG.
The consortium plans to develop a final model based on HyperCLOVA X and ExaOne, featuring a 700 billion parameter expert mixture (MoE) structure. It will develop two independent models—one focused on defense capabilities and the other on attack capabilities—to address the entire cycle of cyber security threats. The performance of the developed models will be objectively verified through two internationally recognized benchmarks.
Data for training will be sourced on a large scale, utilizing approximately 830 terabytes of high-quality real data that would be difficult for a single institution to gather. This data will include operational data from Naver Cloud and LG CNS, as well as data from key national infrastructure and major industries such as Korea Electric Power Corporation, Korea Hydro & Nuclear Power, Financial Security Institute, KISTI, and LG Uplus.
The developed models will undergo continuous enhancement through real-world testing in key national facilities and major industrial sites. Testing will be conducted in seven sectors, including power, finance, science and technology, telecommunications, semiconductors, defense, and aerospace. The consortium also plans to secure specialized operational technology for closed networks to ensure stable operation in security environments with restricted external access.
Kim Yoo-won, CEO of Naver Cloud, stated, “Winning this project reflects the comprehensive recognition of our large-scale AI infrastructure, data, and security expertise. We are committed to creating an AI that protects national critical infrastructure, aiming to develop a representative national security AI model and ecosystem for South Korea.”
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.