Bae, who is also the science minister, made the remarks at a parliamentary audit of his ministry in reply to Rep. Park Jeong-ha of the People Power Party.
The sovereign model project has cleared its second evaluation and is entering a third round, he said.
"Through the second evaluation, South Korea is cementing its standing as the world's third-ranked AI nation," Bae said. He rated the results so far as midrange by global standards. The third phase aims for a model close to the leading open-weight systems used worldwide, he said.
Bae said the bar has risen sharply since last year, as the U.S. government's superintelligence push sets frontier targets beyond earlier goals. He linked the 4.7 trillion won ($3.5 billion) frontier program to the "Mythos shock," a reference to Anthropic's Claude Mythos model.
"The sovereign model will stay on target, but we judged that a frontier AI project is additionally needed to tackle humanity's unsolved problems and respond to cybersecurity threats," Bae said.
He added that current sovereign models would struggle to reach the level industry and the public sector need.
In his opening remarks, Bae named AI data centers, physical AI and semiconductors as three "mega projects." He also said an "AI for All" service, free and with no usage caps, will launch by the end of the year.
AJP Takeaways
- Bae Kyung-hoon said on Oct. 6, 2026, that South Korea will continue its sovereign AI foundation model project and add a separate frontier AI program.
- The sovereign model project has passed its second evaluation. Bae said its results so far are midrange by global standards.
- Bae tied the 4.7 trillion won frontier program to cybersecurity concerns after Anthropic's Claude Mythos, and said an "AI for All" service will launch by year-end.
Copyright ⓒ Aju Press All rights reserved.