South Korea has fired the starting gun on developing a homegrown cybersecurity-focused AI model built on domestic technology.
The move comes as AI-driven hacking threats have become an increasingly real danger worldwide, making it urgent to build the technical capabilities needed to counter them.
The government's flagship initiative to develop an independent AI foundation model — known as the Dokpamo project — is nearing completion. With a security-specialized AI now also in the works, South Korea's push for technological self-reliance is expected to accelerate the formation of a "sovereign AI" ecosystem.
According to industry sources Monday, applications close Wednesday for the "Cybersecurity-Specialized AI Foundation Model Development Project," an initiative led by the Ministry of Science and ICT and the National IT Industry Promotion Agency.
The project aims to pool domestic expertise in AI security and develop a specialized, independent AI model that can be deployed across South Korea's security ecosystem.
The initiative is driven by growing alarm that advanced global AI technologies — including Anthropic's Mithos and OpenAI's latest models — are capable of identifying security vulnerabilities and carrying out unauthorized hacking, making the development of a homegrown AI countermeasure a matter of urgency.
The selected team will receive 256 B200 GPUs (32 nodes) for 10 months to develop the cybersecurity-focused AI foundation model. The government plans to review an interim model at the five-month mark and decide whether to continue support based on the results.
A panel of domestic AI and cybersecurity experts will select the final team, evaluating applicants on three criteria: technical capability and development experience, development goals, and market potential and broader impact.
Industry sources expect SK Telecom, Naver Cloud, LG CNS and Bidraft to form consortiums and submit bids for the project.
Naver Cloud is understood to be forming a consortium with the National Security Research Institute, Theori and Hunesion, along with other AI-related firms and institutions. SK Telecom is partnering with security companies including SK Shielders, AhnLab and Genians. LG CNS is preparing to enter the security-specialized project alongside LG Uplus, LG AI Research and other ICT affiliates, building on its participation in the Modu's AI project. AI startup Bidraft, founded in 2024, is also reviewing whether to join.
If the security-specialized AI model development gets fully under way, the broader AI ecosystem built on homegrown technology is expected to grow more robust.
The Dokpamo project, the government's initiative to develop an independent AI foundation model, has completed its second evaluation round and narrowed the field to three teams. LG AI Research, SK Telecom and Upstage have advanced to the third evaluation round, with the final selection of two teams expected in early next year.
Competition has also intensified for the Modu's AI project, which aims to make AI technology accessible in everyday life. Modu's AI is a government-led initiative to build a free AI chatbot and public AI agent for all citizens using domestically developed AI technology.
Six consortiums have already entered the race for the Modu's AI contract. SK Telecom and KT Corp have each formed their own consortiums as lead companies. The Kakao consortium is led by Kakao and backed by key LG affiliates including LG AI Research, LG Uplus, LG CNS and LG Electronics.
ESTsoft, a first-generation domestic software company known for its Alzip and Alyac products, has also thrown its hat in the ring. MKD and Xenon are participating in the Modu's AI competition without forming separate consortiums.
Selected companies will receive a total of 512 B200 GPUs. The government aims to launch the service by the end of this year following the selection process.
sjpark@heraldcorp.com
