[2026 Audit] Ministry of Science and ICT: "Accelerating AI Drive... Foundation Model Evaluation Criteria Agreed with Companies in Advance"

Bae Kyunghoon Reiterates Commitment to Enhancing AI Competitiveness
Evaluation Criteria for Foundation Model Project “Discussed in Advance With Participating Companies”

At the National Assembly’s Science, ICT, Broadcasting, and Communications Committee audit held on October 6, the Ministry of Science and ICT once again emphasized its commitment to strengthening AI competitiveness, stating that it will not only continue its independent AI foundation model project but also accelerate efforts to foster both frontier AI and physical AI.

Yonhap News Agency

Bae Kyunghoon, Vice Prime Minister and Minister of Science and ICT, reaffirmed that while the independent foundation model project will proceed as originally planned, an additional initiative will be launched to develop frontier AI, particularly to respond to cybersecurity challenges.

Explaining the background of the frontier AI project, estimated at 4.7 trillion won, he said, “After the Mythos shock, we questioned whether we could secure AI capable of responding to cybersecurity threats, and concluded that the current level of our independent foundation model is unlikely to reach the standards required for industrial and public sector AI applications.”

Yonhap News Agency

Meanwhile, when Kim Jangkyum, Assemblyman of the People Power Party, raised concerns over the fairness of the evaluation process for the independent foundation model project, Ryu Jemyoung, 2nd Vice Minister of Science and ICT, responded, “The evaluation criteria and scoring were thoroughly discussed in advance with all participating companies, ensuring mutual understanding before the assessment took place.”

Assemblyman Kim pointed out both the financial limitations of the companies developing datasets used to evaluate the project and procedural issues in the assessment. For the mathematics dataset, the lead company reportedly had annual sales of only 150 million won and eight employees, while the knowledge dataset lead company had annual sales below 1 billion won and fewer than 20 staff.

He also noted that 33 out of the 49 experts serving on the first and second AI specialist user evaluation panels overlapped, and raised fairness concerns over the fact that in the expert evaluation category, which carried the highest weight, LG AI Research secured first place, overturning Motif Technologies, which achieved the highest AI benchmark score.

Additional issues raised included the fact that conflict-of-interest controls were implemented solely through written pledges, and that certain AI model development targets were lowered.

Vice Minister Ryu explained, “We held one-on-one discussions regarding detailed scoring standards, methods, and procedures for anonymity, and ultimately made the rules of the game official in a meeting attended by all participating companies. We also provided guidance on specific details related to the evaluation process both before and after the assessments.”

This content was produced with the assistance of AI translation services.