AI Models Become Strategic Assets Amid U.S.-China Tensions

The U.S. government has tightened export controls on Anthropic's next-generation models, 'Preamble 5' and 'Mythos 5,' amid concerns over the potential leakage of AI models to China through South Korea. Lawmakers in the U.S. are increasingly focused on the security implications of AI model leakage related to South Korea.

According to U.S. political sources on June 21, the White House has identified that the Chinese government has tasked its domestic AI companies with acquiring information on U.S. AI models as a national priority, promising rewards for successful efforts. This indicates that China is not only aiming to enhance its own models but is also pursuing technical details of U.S. frontier models as a strategic national objective.

There are indications that South Korean small and medium-sized AI infrastructure companies have become targets in this process. Multiple AI infrastructure firms have reported that Chinese entities have requested certain rights over AI data centers operating in South Korea, citing investment proposals. Analysts suggest this may be an attempt to gain access to U.S. AI models.

China's interest coincides with a funding crunch in South Korea's AI data center expansion. Operating AI data centers requires a massive supply of GPUs, but many domestic infrastructure companies are struggling with cash flow. U.S. political observers believe that large-scale Chinese capital is stepping in to fill this gap, thereby positioning South Korea as a base for competition with the U.S. in AI models.

However, this situation cannot be viewed merely as an issue of capital influx. South Korea is seen as strategically positioned in the geopolitical landscape, serving as a battleground for U.S.-China cyber warfare. As high-performance AI models are elevated to strategic assets across military and industrial sectors, the competition for AI supremacy between the U.S. and China is intensifying within South Korea.

In this context, South Korean telecommunications company SKT's participation as a partner in Anthropic's 'Project Glasswing' has raised concerns about potential ties to China.

SKT has clarified that its involvement with Project Glasswing is entirely unrelated to any attempts by China to access Mythos 5. The partnership is based on a separate collaboration with Anthropic, differing from the allegations raised.

Lobbyists in Washington, D.C., also sense that the White House is not specifically targeting SKT. Instead, the concern appears to be focused on the potential for Chinese infiltration into the broader South Korean AI infrastructure ecosystem.

Industry insiders view this issue as a reflection of the structural vulnerabilities within South Korea's AI sector. The surge in GPU demand due to the expansion of AI data centers, coupled with the funding limitations faced by domestic infrastructure companies, has created an environment of increased reliance on foreign capital. There are calls for the government to establish a support system for AI infrastructure investment and to implement separate verification processes for technology access when foreign capital is involved.

An IT industry representative stated, "Global capital is flocking to the U.S., while domestic capital is effectively stagnant, making the influx of Chinese capital into costly AI infrastructure inevitable. Ultimately, as reliance on Chinese capital for AI infrastructure increases, the U.S. will likely heighten its vigilance toward South Korea."

* This article has been translated by AI.