A dispute over the effectiveness of the government's sovereign AI foundation model program took center stage at this year's parliamentary audit.
At a National Assembly Science, ICT, Broadcasting and Communications Committee audit of the Ministry of Science and ICT on the morning of Oct. 6, lawmakers pressed officials on the program's results and on how the government decided to launch a separate frontier AI initiative. Rep. Park Jung-ha of the People Power Party said the program has produced too little given how heavily the government promoted it, and asked whether the government had offered nothing more than a rosy blueprint. He questioned in particular whether the large-scale frontier AI project the government plans to begin next year overlaps with the existing program and signals a shift in policy direction while the foundation model effort is still under way.
Park also challenged the decision-making process behind the frontier AI project. Government funding alone comes to about 4.7 trillion won ($3.3 billion), and because the government would hold a 30% stake, the total project would reach roughly 16 trillion won, he said, yet the process leading to the plan has not been sufficiently disclosed. "There is no way to confirm where and how this was decided," Park said. "It looks like it fell out of the sky."
In response, Deputy Prime Minister and Science Minister Bae Kyung-hoon said the ministry will continue the foundation model program as planned. "The sovereign AI foundation model was developed to spread AI into existing industries and the public sector and to improve its usability, and it has been pursued with the goal of developing a world-class model," Bae said. "We have completed the second round of evaluation and have entered the third stage."
Bae added that the benchmarks set last year and those applying this year are now markedly different. "With the U.S. government announcing superintelligence, we are drawing up plans for a frontier model that goes several times beyond the level we had originally targeted," he said. Competition in AI technology has intensified faster than expected, he said, making an additional development project at a higher level than the existing program necessary.
In the third stage of the foundation model program, the government plans to develop a model approaching the level of leading global open-weight models, while the models secured through the second evaluation will be rolled out in earnest across the public and industrial sectors. Bae stressed that the program is building an ecosystem that includes not only large companies but also startups and other domestic AI firms.
Bae cited the recent "Mythos shock" and cybersecurity threats as the reasons for adding the frontier AI effort. "When we assessed whether we could secure AI capable of responding to cybersecurity threats after the Mythos shock, we concluded that the current level of the sovereign AI foundation model would make that difficult to achieve," he said. "Frontier AI is not a project that suddenly fell out of the sky."
He added that the ministry needs to pursue frontier AI to address humanity's unsolved problems and cybersecurity challenges that existing AI cannot handle, while keeping the foundation model program on its original course — making clear that the two will be pursued as separate projects.
