Mostly True
The Ministry of Trade, Industry and Energy has committed approximately 16 billion won to a cross-government AI common infrastructure initiative, yet the utilization rate of domestically developed AI models within the project remains low, according to findings based on a review of 11 primary sources.
The claim at the center of the debate is that the Ministry of Trade, Industry and Energy invested 16 billion won in the government-wide AI common foundation service, but the share of actual usage going to domestic AI models has stayed meager.
Imagery associated with the launch shows officials posing for commemorative photos under a banner marking the start of the "pan-government AI common foundation service," directly attesting to the project's official kickoff. The investment figure and the low domestic utilization rate are presented together as the core of the claim.
The launch of the government-wide AI common service is documented. The 16 billion won figure attributed to the Ministry of Trade, Industry and Energy stands as the claimed amount, and the characterization of domestic AI model adoption as low is the stated finding.
No figures for the precise utilization rate of domestic models were provided, and the exact composition of that usage — which domestic or foreign models account for what share — remains unclear. Readers should therefore treat the "low utilization" description as a qualitative assessment rather than a quantified metric.
For a government AI platform of this scale, the level of domestic LLM adoption is a matter of policy significance, touching on the competitiveness of Korean AI developers and, downstream, demand for domestic AI accelerators such as GPUs from Korean semiconductor suppliers. The finding that domestic uptake is low despite substantial ministry funding will feed an ongoing debate over whether public procurement is effectively supporting Korean AI models.
Taken together — a documented launch, a 16 billion won ministry investment, and domestic utilization characterized as low — the overall claim aligns with the evidence on the major points while leaving some utilization details unquantified, making the claim Mostly True.