Korea's independent AI model project: second-round results and an agent for every citizen

The key claims in the second-round results of the independent AI foundation model project, as reported by the Korean government, are that only the US, China and Korea have made Epoch AI's Notable AI Models list, and that seven Korean companies have scores measured by Artificial Analysis. Four consortia took part — SK, LG AI Research (750 billion parameters), Upstage (No. 9 by daily usage on OpenRouter) and Motif (No. 6 among open models) — followed by figures such as legal AI cutting litigation preparation by 45.7% and new-material discovery shrinking from 22 months to 1 day. However, Korean models have 300 billion to 750 billion parameters, a gap from Big Tech frontier models in the trillions, so a division of roles was proposed: startups build small industry-specific models, while large companies take on the frontier together with the government. The one-agent-per-citizen plan, which would assign every citizen an AI assistant that finds and applies for welfare services on its own, faces the challenges of data integration across ministries and personal data leaks; the balance point is a basic version for which the state takes responsibility using minimal information, with anything beyond that left as a consent-based option. All figures are claims made by those presenting the results, so they need a caveat when cited.

This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.