NAVER Cloud Consortium to Develop AI Models Tailored to Korea’s Cybersecurity Environment
NAVER Cloud Consortium to Develop AI Models Tailored to Korea’s Cybersecurity Environment
NAVER Cloud Consortium to Develop AI Models Tailored to Korea’s Cybersecurity Environment
- Consortium deploys 4,000 GPUs ahead of project launch to develop two of its own cybersecurity models, each with approximately 700 billion parameters, based on HyperCLOVA X and EXAONE
- Planned field trials across seven strategic industries and critical national facilities aim to expand Korea’s cybersecurity AI ecosystem
September 3, 2026
The NAVER Cloud consortium announced that it has been selected to carry out the government’s specialized AI foundation model development project in cybersecurity and is beginning full-scale development. The consortium brings together 33 organizations from industry, academia, research and the public sector to build AI models optimized for Korea’s cybersecurity environment. It plans to test and refine the models at critical national facilities and across major industries to strengthen Korea’s competitiveness in cybersecurity AI.
The consortium began pretraining its cybersecurity-specific AI models before the project’s official launch, deploying 4,000 GPUs from its existing resources. NAVER Cloud is contributing 4,000 B200 GPUs and LG is contributing 256 H200 GPUs, in addition to the 256 GPUs provided by the government. This substantial upfront investment is intended to significantly shorten development time, allowing the consortium to focus the time saved on improving model performance and conducting field trials to maximize the quality of the final models.
Using this large-scale infrastructure, the consortium will develop hyperscale AI models specialized in cybersecurity, based on HyperCLOVA X and EXAONE. Each model will ultimately have approximately 700 billion parameters and a mixture-of-experts (MoE) architecture. The consortium will develop two models of its own in parallel: a HyperCLOVA X-based model with enhanced defensive capabilities and an EXAONE-based model with enhanced offensive capabilities. Together, they will cover the full attack-and-defense cycle. Performance will be rigorously evaluated using two internationally recognized benchmarks to provide an objective assessment.
The training dataset, a key determinant of model performance, will comprise approximately 830 TB of high-quality, real-world data that would be difficult for any single organization to obtain. The dataset combines operational data from NAVER Cloud and LG CNS with key data from critical infrastructure and major industries, including KEPCO KDN, Korea Hydro & Nuclear Power, the Financial Security Institute, the Korea Institute of Science and Technology Information (KISTI), and LG Uplus. The consortium will apply data refinement techniques spanning threat intelligence collection, labeling, standardization and validation to train the models to accurately understand real-world threat environments.
The models will be continuously refined through field trials at critical national facilities and across major industries. Trials will cover seven key sectors: power, finance, science and technology, telecommunications, semiconductors, defense, and aerospace. The consortium also plans to develop specialized technology for operating the models on closed networks, enabling reliable use in secure environments where access to external networks is restricted.
The consortium plans to release the models as open source for commercial use, extending the project’s benefits across Korea’s cybersecurity ecosystem. It will offer small and medium-sized enterprises free assessments to identify vulnerabilities and detect cyberattacks, while helping Korean cybersecurity companies commercialize AI services through the NAVER Cloud Marketplace. These efforts aim to build a national cybersecurity AI ecosystem in which development and adoption reinforce each other.
“Our selection for this project reflects comprehensive recognition of NAVER Cloud consortium’s combined strengths in large-scale AI infrastructure, data, and cybersecurity expertise,” said Kim Yu-won, CEO of NAVER Cloud. “We have a responsibility to build AI that protects critical national infrastructure using our own technology. With that responsibility in mind, we will work to develop a flagship national cybersecurity AI model and ecosystem for Korea.” (End)
