Korea outlines 3.5 trillion won plan to develop frontier AI

The government plans to develop a frontier AI model that would require an investment of more than 3.5 trillion won ($2.36 billion), Science and ICT Minister Bae Kyung-hoon said Monday.

Korea’s state-led AI ambitions were laid out during a press briefing at the Government Complex Sejong, with Bae explaining how the government is preparing to counter the most advanced models emerging from the United States and China. 

The California-based AI firm Anthropic released Mythos 5 in April, and the Beijing-based firm Moonshot AI unveiled Kimi K3 on Friday. 

As part of the response, Bae outlined a “two-track strategy” that would pursue both publicly available AI models and frontier-level models in parallel.

Under the plan, the government's ongoing sovereign foundation model project will continue to focus on developing AI models for broad use across the public sector. It also envisions a separate initiative to build a cutting-edge model.

"Developing a frontier model would require about 3.5 trillion won just to purchase 10,000 of Nvidia's new Vera Rubin chips, which are scheduled for release in the second half of the year," Bae said. "We are in the process of persuading relevant ministries and fiscal authorities."

As a first step, the government plans to develop a security-focused AI model capable of identifying vulnerabilities by the end of the year to deploy in the public sector.

Bae also rejected the idea of relying on foreign open-source AI models as an alternative.

Unlike the United States, which is moving toward a more restrictive AI policy by requiring a government review before AI models are released, China has promoted a more open AI ecosystem by making its models open source.

"China could also switch to a closed model at any time once it reaches a certain level," Bae said.

He stressed that Korea still needs to strengthen its own AI capabilities regardless of the availability of open-source models.

BY KIM MIN-JEONG [[email protected]]