Korea’s AI ambitions at crossroads amid US push for tech dominance

U.S. President Donald Trump holds a signed executive order during an artificial intelligence summit in Washington, July 23. AP-Yonhap

In the wake of the U.S. AI Action Plan, which prioritizes domestically developed technologies, Korea’s ambitious artificial intelligence strategy finds itself at a critical juncture.

Korea has launched a project to develop its own proprietary AI foundation model, aiming for technological self-reliance by building large language and multimodal models entirely from scratch. The goal is to reach at least 95 percent of the performance of leading global AI systems, an ambitious step toward achieving technological autonomy in AI.

Following an initial application review, 10 consortia — including ones led by Naver Cloud, LG AI Research, Upstage and KAIST — have been shortlisted. This group will be narrowed down to five finalists by early August, with the government providing support in graphics processing units, data and personnel. By 2027, two final teams will be selected.

However, Korea’s plans to champion sovereign AI face questions as Washington launches its own strategy in a clear play to consolidate the global AI ecosystem under its standards and supply chains.

U.S. President Donald Trump signed AI-focused executive orders last week with plans to accelerate innovation, build American AI infrastructure and take a leading role in the global market for AI. The plan also seeks to export “full-stack” American AI, including hardware, models and standards, to allied nations while blocking adversary access and tightening export controls.

For Korea, the U.S. move raises concerns that any locally developed AI could be shut out of key export markets or collaborative opportunities unless it integrates with the American stack.

The Korea Software Industry Association (KOSA) warned in its recent report that if core AI models and platforms are built around U.S. standards, Korean companies may be relegated to the role of “lower-level suppliers” developing apps for foreign platforms.

However, experts caution that it is too early to determine the fate of the project, suggesting a balanced approach that leverages Korea’s unique strengths while maintaining strategic flexibility.

LG AI Research co-President Lee Hong-lak speaks during an LG AI Talk Concert event at LG Science Park in western Seoul, July 22. Courtesy of LG AI Research

Professor Choi Byeong-ho of Korea University’s Human-inspired AI Research said that if Korea adopts a fundamentally different approach rather than following in the footsteps of the U.S. or China, it can gain a competitive edge despite pressures from the two countries.

“At the end of the day, the real purpose behind building up AI infrastructure or models, considered inputs, is to deliver competitive AI services, which are outputs … Right now, all this input is part of a game of economies of scale, and we just can’t win that fight. That’s why we have to turn the equation around,” he said.

“We should optimize everything with the end goal of building export-ready, high-quality AI products. That’s how we can carve out our own path.”

Noting that Korea should approach AI in the way it approached semiconductors, he suggested that focusing on lightweight models could be a way for Korea to gain a competitive advantage.

“Even if there are some trade-offs, what matters is having a model that can deliver optimized, high-quality output. If we can come up with a well-optimized lightweight model, we could really pack a punch,” he said.

“Our strategy should be like semiconductors: always export-focused. If we cannot break into the U.S., then we go for Southeast Asia, Europe or wherever we can. Korea has always survived by diversifying its export markets, so we should go as broad as possible in our push outward.”

Gachon University business professor Jeon Seong-min explained Korean consumers’ tendency for fast-changing and unique demand could help the country to come up with differentiating outputs.

“It would not be right to avoid trying, just because we are worried about potential threats or possible trade disputes. In the process of developing our own models, we could succeed or fail, but there is so much to learn along the way,” he said.

“If we can achieve something in an area where we already excel, that has the potential to become mainstream.”

However, from the players’ perspective, the critical issue is the outcome of the upcoming trade negotiations, which currently face an Aug. 1 deadline, according to an industry insider.

“Right now, the most important thing is not funding; it is trade. If trade issues are not resolved, what is the point of building the world’s best AI?” he said.

“The current trade situation encompasses everything — semiconductors, electronics, all bundled together. Unless that is resolved, the industry cannot move forward. We need those barriers removed, so we can truly compete to build world-class foundation models.”