Naver Cloud has won a competitive government bid to lead the development of a cybersecurity-specific AI foundation model, beating out a consortium led by SK Telecom. The Ministry of Science and ICT (MSIT) announced the selection on September 3, placing Naver Cloud at the head of a large-scale industry-academia-research project funded by the state.
The decision reverses an earlier outcome between the two companies. SK Telecom passed both evaluation stages of the government's separate Sovereign AI Foundation Model project, while Naver Cloud was eliminated in the first stage in January. In the security-focused project, the result flipped: Naver Cloud secured the final contractor position over SK Telecom.
How the evaluation played out
MSIT said external experts in AI and cybersecurity assessed each consortium's business proposal and presentation. Evaluators scored technology capabilities, development experience, strategy, goal quality, feasibility, market potential, and broader impact. The Naver Cloud consortium drew favorable marks for building a large-scale collaboration framework and for its plan to connect and integrate multiple AI agents.
The rival SK Telecom consortium brought significant AI development credentials to the table. SKT and Upstage - both teams that passed the second stage of the Sovereign AI Foundation Model project - joined forces. The consortium also included nine cybersecurity firms: SK Shieldus, SECUI, AhnLab, IGLOO, RaonSecure, SmartM2M, AIM Intelligence, Genians, and PIOLINK. Korea University, Soongsil University's Industry-Academic Cooperation Foundation, and the Korea Information Security Industry Association rounded out the 14-organization group.
Naver Cloud's consortium matched that depth on the security side. LG CNS, LG AI Research, and LG Uplus joined as core partners, alongside cybersecurity and AI companies S2W, ESTsecurity, and JiranSecurity. The Financial Security Institute, KEPCO KDN, Korea Hydro & Nuclear Power, the Korea Institute of Science and Technology Information, and several universities and research institutions also signed on.
A two-model approach with agent architecture
Naver Cloud proposed developing two models in parallel: one specialized for defensive cybersecurity tasks, the other for offensive capabilities. The defensive and offensive models will build on Naver's HyperCLOVA X and LG AI Research's EXAONE respectively. The end target is a cybersecurity model with a 700-billion-parameter mixture-of-experts architecture.
The government emphasized the consortium's use of "multiple agents and a harness" in its announcement. The agents are AI systems assigned to distinct cybersecurity functions - threat detection, vulnerability analysis, attack analysis, and response. The harness serves as an integrated execution framework that connects and manages those agents, the underlying AI models, and external security tools.
MSIT did not release individual evaluation scores or the margin between the two consortiums, leaving the specific deciding factors unclear. SK Telecom said it will continue pursuing cybersecurity AI development independently and plans to discuss future cooperation with the organizations that joined its consortium.
Computing resources and timeline
The Naver Cloud consortium will receive access to 256 Nvidia B200 GPUs for 10 months starting this month. The government will provide the GPUs for an initial five-month period. Support for the remaining five months depends on the results of an interim evaluation.
Why this matters for government professionals
This project signals a concrete government investment in domain-specific AI for national security infrastructure. For agencies and public-sector IT leaders, the selection process offers a reference point for evaluating AI consortiums: the winning bid succeeded on the strength of its collaboration framework and its technical architecture for coordinating multiple specialized AI agents, not just raw model performance. The split evaluation outcome between this project and the Sovereign AI project also shows that government panels weigh domain expertise and deployment strategy differently depending on the mission. Professionals involved in AI for Cybersecurity Analysts or procurement for AI for Government can study this case as a worked example of how technical proposals are assessed when security is the primary requirement.
Your membership also unlocks:
