SK Telecom and NAVER Cloud are squaring off for the role of building South Korea's flagship cybersecurity artificial intelligence model. Each company has assembled a large consortium and submitted a bid for the government's project to develop a cybersecurity-focused AI foundation model. Notably, SK Telecom, Upstage and the LG AI Research Institute — the three finalists in the country's sovereign AI foundation model program — have split into opposing camps, setting up a contest of pride among Korea's leading AI firms.
Growing Global Security Threats Prompt Government Push for Korean-Language Model
SK Telecom and NAVER Cloud said on the 27th that each had submitted a proposal for the government's cybersecurity-focused AI foundation model development project, for which bids were accepted from the 22nd of last month through the 26th of this month. The Ministry of Science and ICT plans to review the eligibility of the submitted documents and select a project team in early September after external expert presentations and evaluations.
The global security industry has recently moved toward combining security-focused AI models with AI agents to automate not only threat detection but also blocking and remediation. As cyberattacks grow more complex and faster, the conventional approach of having humans analyze and respond to each threat has run into limits. In Korea, however, there is still a shortage of dedicated AI models trained on Korean-language security data and capable of understanding domestic regulatory frameworks. For that reason, the government is pursuing the cybersecurity-focused AI foundation model project to develop a model tailored to Korea's cyber threat environment and Korean-language security data, and to apply it in actual security operations.
The two consortiums that entered the competition both aim to develop AI tailored to Korea's security environment, but they have put forward different strengths. SK Telecom is emphasizing an agentic security AI that continuously learns from actual security monitoring data and automates the full chain from threat detection to analysis, response and remediation. NAVER Cloud is highlighting a security AI that can operate reliably in air-gapped and on-premises environments, keeping sensitive data from national critical facilities from leaving the premises.
SK Telecom Teams With Upstage on Agentic Security AI
SK Telecom has joined forces with Upstage, its rival in the sovereign AI model program. The two companies plan to develop a security-focused model based on the A.X K and Solar Open series they have each developed. They intend to combine the model with AI agents so that it does not merely detect cyberattacks and alert security staff, but analyzes and assesses threats and carries out responses and remediation automatically.
Another strength the SK Telecom camp points to is a cycle running from AI model development to training on actual security data, field validation, and evaluation and retraining. Participating security firms will supply data accumulated in real monitoring and security operations for model training, and the resulting model will be tested again in each company's commercial products and monitoring environments. The test results will then feed back into model training to steadily improve field response capability.
For the project, SK Telecom and Upstage will handle model development, while nine security firms — SK Shieldus, SECUI, AhnLab, Igloo Corporation, Raonsecure, SmartM2M, Aim Intelligence, Genians and Piolink — will provide actual security data and take part in field validation. Korea University and Soongsil University will verify the model's performance and safety, and the Korea Information Security Industry Association will handle the spread of the results across the industry. Including the lead company, the SK Telecom consortium comprises 14 organizations.
NAVER Cloud and LG Form One Team, Betting on Air-Gapped and On-Premises Systems
On the other side, LG CNS, the LG AI Research Institute — another sovereign AI model finalist — and LG Uplus have lined up around NAVER Cloud. NAVER Cloud has assembled a consortium of 33 organizations, including the lead company. Alongside security firms such as Logpresso, Sands Lab, S2W, ESTsecurity and Theori Korea, institutions tied to national critical infrastructure have joined, including Korea Hydro & Nuclear Power, KEPCO KDN, Korea Aerospace Industries (KAI), the Financial Security Institute and the Korea Institute of Science and Technology Information (KISTI). Universities including Seoul National University, KAIST, POSTECH and Korea University are also taking part.
NAVER Cloud is presenting its full-stack capability, spanning AI models, infrastructure, security and data, as its differentiator. In particular, NAVER Cloud has experience building and commercially operating foundation models directly in air-gapped and on-premises environments separated from the external internet, in fields requiring high levels of security such as nuclear power and finance. The company aims to develop a model suited to national critical facilities that must use AI internally without sending sensitive data, such as attack logs and network information, outside.
LG CNS will contribute specialized cybersecurity units including red team, blue team and AI security team, along with AI transformation experience across various industries. Telecommunications network data from LG Uplus and others will also be used to refine the model. The plan is to combine security firms with attack and detection technology, public institutions holding national-level threat data, and universities researching AI safety, building a system capable of handling everything from model development to data acquisition, field testing and verification within Korea.
Government Aims to Reduce Reliance on Foreign AI Models, Weighs Open-Source Release
The government plans to use the project to strengthen the competitiveness of frontier-class AI models and to secure a cybersecurity-focused AI model suited to the domestic environment as quickly as possible. The selected model will be made available to the security ecosystem in open-source form, with support for further refinement planned through 2027.
Bae Kyung-hoon, deputy prime minister and minister of science and ICT, wrote recently on social media that AI has moved beyond being a support tool for cybersecurity and into a stage where it changes the very rules of attack and defense, adding that the core of an AI security model ultimately comes down to the performance of the foundation model.
