Summary
As the U.S. administration plays the export-control card on Mythos, the top-tier model from AI company Anthropic, a roadblock has emerged for Glasswing, the global AI cooperation project that Korea had been pushing to join. The Korean government is taking a cautious stance, saying it is still assessing the situation. The move is read as a signal that the battle for tech supremacy over access to frontier AI models is now extending even to allied nations.
How It Unfolded
Until now, the U.S. has concentrated on export controls covering advanced semiconductors and AI chips, but this time it has begun weighing controls on the top-tier large language model itself rather than the chips. Anthropic's Mythos is regarded as a model with best-in-class performance in reasoning, coding, and large-scale analysis, and the step is interpreted as an effort to restrict the cross-border movement of such frontier models' weights and core technology.
The crux of the problem is that this control has emerged as a wild card just as Korea was in the process of joining Glasswing, the U.S.-led multinational AI cooperation body. With the eligibility and terms for accessing the top-tier model now opaque, the very timeline for participation by Korean companies and research institutions has become uncertain. Government authorities, refraining from an official statement, are at the stage of verifying Washington's intentions and the scope of application.
The key issue is whether the controls target the model in its entirety or a specific high-performance version, and what exceptions might apply to allied nations. The broader the scope of application, the more the entire structure of Korea's AI ecosystem — which has relied on overseas frontier models — could be shaken.
Structural Background
To date, a large share of Korean industry and startups have built AI services by renting U.S.-made top-tier models via the cloud. Only a handful possess their own ultra-large-scale models. If the U.S. widens the front to model-level controls, companies without their own model capabilities will find their core infrastructure at the mercy of external policy.
Paradoxically, this could become a catalyst that stimulates national-level AI sovereignty and demand for sovereign AI. As the drive to bring data, models, and infrastructure under domestic control intensifies, the strategic value of camps that own proprietary models and homegrown infrastructure comes to the fore.
Impact on Stocks and Sectors
- Naver: With its own ultra-large-scale model, HyperCLOVA X, it could emerge as the top candidate to benefit from sovereign AI alternative demand.
- Kakao: Backed by its own model and platform, it stands to gain spillover benefits as the homegrown AI ecosystem expands.
- Samsung SDS and SK Telecom: Should demand for localization rise in enterprise AI solutions and data center businesses, their order opportunities would grow.
- Samsung Electronics and SK Hynix: If the AI-semiconductor control stance spreads into the model domain, the double-edged factor of supply-chain uncertainty comes into focus.
- Domestic AI infrastructure and cloud industry sector: A divergence could play out, with companies heavily dependent on overseas models facing a short-term shock and the homegrown-alternative camp seeing opportunity.
Bullish vs. Bearish Scenarios
In the bullish scenario, as the controls materialize, the government and companies accelerate investment in homegrown models and infrastructure, policy support and budgets become concentrated, and the medium- to long-term growth potential of Korea's AI sector bellwethers is re-rated. A structural benefit is possible, with capital and talent converging on the handful of companies that possess their own capabilities.
In the bearish scenario, access to top-tier models is blocked, the quality competitiveness of Korean AI services retreats in the short term, and the risk of being sidelined in global cooperation grows. While the scope of the controls remains ambiguous, investor sentiment may contract and the volatility of related stocks may widen.
Investor Action Points
- Avoid overreacting to short-term spikes and drops until the official scope of the controls and whether allied-nation exceptions apply are announced.
- Take a differentiated approach, distinguishing the homegrown AI camp that owns its own ultra-large-scale models and infrastructure from companies heavily dependent on overseas models.
- Track the government's sovereign AI policy and budget announcements as a momentum indicator.
- Keep the question of whether the AI-semiconductor supply-chain control trend spreads into the model domain as a risk-check item.
This article is content automatically summarized and analyzed based on the original news report. View original (Yonhap News, Industry)
