South Korea is pressing ahead with its existing national AI foundation-model project while planning a separate frontier-model programme worth 4.7 trillion won ($3.5 billion). The two-track plan raises a practical question for the teams competing for public support: what will each programme deliver that the other cannot?
Watch Desk analysis
What happened
The Korea Times reports that the existing project will enter a third-stage evaluation, with two finalists expected to be selected in February. The separate frontier AI programme is due to begin in March 2027, with 4.7 trillion won allocated to it.
Science Minister Bae Kyung-hoon says the existing project is continuing, and that the two tracks serve different needs. The article says the government is considering changes to the scale and form of support for the existing project’s finalists, as well as possible incentives or priority for finalists that join the frontier programme. LG AI Research, SK Telecom and Upstage are the three teams still competing.
The article also reports expert concerns that the programmes could overlap, duplicate spending or leave the purpose of the existing project unclear. One proposed alternative is to build on existing models rather than start from scratch. These are reported policy questions, not a settled government decision.
Why it matters
This is a significant public investment in AI development, but the distinction between the two tracks is still the key to judging it. The article says the existing project originally aimed to reach about 95 per cent of leading global models’ performance. Officials now argue that a separate effort is needed to pursue frontier capabilities, while the current project continues towards its own goals.
That leaves a consequential choice about where money, expertise and future government demand will go. The government is also considering an AI adoption track to expand the use of sovereign models in industry and public services, so the programmes’ eventual value may depend on whether models find distinct uses, not simply which one tops a leaderboard.
Our read
Keeping both tracks alive could preserve the skills and models already developed while giving South Korea another route to frontier systems. But “different needs” needs to become a clear brief: without defined goals, timelines and support terms, two programmes can look less like a strategy and more like two queues for the same pot of talent.
The immediate news is that the existing competition continues and a separate, much larger frontier programme is planned. The harder test will be whether the government can explain how their purposes differ and show what public money is buying.
What to watch
- Which two teams advance in the existing project’s February evaluation.
- How the government defines the frontier programme’s goals and timetable before its planned March 2027 start.
- Whether finalist support changes, and what incentives or priority are offered for joining both tracks.
- Whether the proposed AI adoption track leads to clearly identified uses across industry and public services.
Discussion spark: Should South Korea keep funding both AI tracks, or concentrate public support on the programme with the clearest route to frontier-level models?
Sources and evidence
- Source update (11 October 2026, 05:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.