Korea’s Two-Track AI Strategy Raises Questions Over Funding and Focus

South Korea is moving ahead with two government-backed artificial intelligence programs, a strategy intended to support both practical deployment and cutting-edge research but one that is also raising questions about overlapping goals and resources.

The Ministry of Science and ICT plans to continue its national AI foundation-model project through a third-stage evaluation. Two finalists are expected to be selected in February. At the same time, the government is preparing a separate frontier AI initiative scheduled to begin in March 2027, with 4.7 trillion won allocated in the proposed budget.

The existing program was launched to build domestic foundation-model capabilities and encourage wider use of Korean AI systems. Its original benchmark was to reach roughly 95 percent of the performance of leading global models. Rapid advances by international developers, however, have changed the competitive landscape and made that target harder to interpret.

Deputy Prime Minister and Science Minister Bae Kyung-hoon said the foundation-model effort would focus on expanding AI adoption in industry and the public sector and bringing Korean models closer to widely used open-weight systems. He presented the frontier project as a separate effort aimed at more ambitious technical challenges, including problems that existing systems cannot solve and emerging cybersecurity threats.

Officials argue that the two programs can serve different layers of the AI market. Lightweight and industry-specific models may remain valuable even if they do not match the largest systems, while frontier models can target the most demanding research and strategic applications. The government has also indicated that the two finalists in the current competition will continue to receive support, although the scale and form of that assistance remain under consideration.

The unresolved issue is how clearly those roles will be separated. Experts have warned that a successful frontier model could become the preferred option for domestic companies and public institutions, weakening demand for models produced under the existing program. They have also cautioned that the two tracks could duplicate spending on data, computing resources and development work unless the government defines distinct objectives and timelines.

Three teams led by LG AI Research, SK Telecom and Upstage remain in the current competition. The government is considering incentives or priority for the eventual finalists if they join the frontier program, alongside a separate adoption track intended to spread sovereign AI systems across industries and public services. For participating companies, the approach could create a pathway from model development to deployment, but it could also complicate investment and staffing decisions until the government clarifies how the programs will interact.