Korea's Ministry of Science and ICT disclosed second-round evaluation scores for its program to develop a domestically built AI foundation model. The step followed an objection from Motif, a participant eliminated in that round, which questioned the fairness of the assessment.

According to the released figures, SK Telecom finished first, while the objecting company placed fourth, below the cutoff. The ministry published the composition of the review panel, the scoring rubric and item-by-item marks, saying the process contained no procedural flaw.

How the program works

The program aims to secure a home-grown base model rather than depend on large overseas systems. Consortia are narrowed through staged evaluations, with computing resources and budget concentrated on the survivors. Elimination means losing access to those resources, which makes the contest fierce.

Objections are a recurring risk in that design. The more qualitative judgment an evaluation contains, the heavier the duty to explain the outcome.

What disclosure leaves behind

Publishing scores ends a dispute quickly, but it also exposes each participant's technical standing. The episode raises the question of how state-led R&D programs can deliver transparency and protect participants at the same time.

The ministry said it would set out criteria and schedules more explicitly in advance for the remaining stages.