SEOUL — LG Group is making a strategic push to participate in the South Korean government’s national project to develop a cybersecurity-specialized Artificial Intelligence (AI) foundation model. Built upon LG AI Research’s proprietary hyper-scale AI model, EXAONE, LG’s entry escalates competition against tech giants like Naver Cloud and SK Telecom, both of which are aggressively forming their own rival consortia.

According to industry sources on August 12, key affiliates including LG CNS, LG AI Research, and LG Uplus are actively finalizing a consortium to bid for the "Specialized AI Foundation Model Development (Cybersecurity Sector)" project, spearheaded by the Ministry of Science and ICT (MSIT).

LG plans to wrap up consortium agreements with cybersecurity firms and academic institutions shortly before submitting its formal application. While the involvement of LG affiliates is confirmed, specific third-party partners have not yet been publicly disclosed.

Driving Cybersecurity Independence via Tailored AI

Unlike general-purpose LLMs (Large Language Models), the core goal of this government initiative is to create a tailored AI foundation model specifically optimized for national and enterprise cybersecurity tasks. To empower the winning consortium, the South Korean government will supply 256 units of NVIDIA's next-generation B200 GPUs for a period of 10 months.

The LG Consortium intends to leverage EXAONE as its core architecture, fine-tuning it with specialized security datasets to drastically improve its autonomous threat detection, vulnerability analysis, and incident response capabilities. LG will rely heavily on its internal enterprise security expertise to guide model development and rigorous validation.

Clear Strategic Roles Among LG Affiliates

Within the consortium, tasks are divided to maximize domain expertise:

LG AI Research & LG CNS: Jointly fine-tune EXAONE using high-value security datasets to maximize operational precision for cybersecurity workflows.
LG CNS: Oversees the data pipeline—including data cleaning, labeling, standardization, and processing—as well as comprehensive performance and safety verification of the finalized model.
LG Uplus: Supplies critical telemetry and security data necessary for training while conducting real-world empirical validation and testing across telecom network infrastructures.
"Through this project, we aim to secure a cybersecurity-specialized AI foundation model and build an environment where domestic security companies and public institutions can leverage it to create diverse security services," said an official from the LG Consortium. "Our goal is to contribute significantly to vitalizing South Korea's broader AI security ecosystem."
As major domestic cloud and telecommunication providers vie for the government's high-performance GPU support, LG’s joining underscores the accelerating race across corporate Korea to dominate specialized domain AI solutions.