One Suspected China Link. One Korean Telecom. Anthropic’s Entire Flagship AI — Offline

Washington's paranoia about Chinese AI access has reached a new threshold: an unproven suspicion was enough to take a frontier model offline worldwide

A $100 Million Partner, Suddenly a Liability

In early June, Anthropic expanded access to its most advanced AI system — Claude Mythos — to roughly 150 organisations across more than fifteen countries, as part of a security initiative called Project Glasswing. Organisations reported to have received access in South Korea included the Korea Internet & Security Agency, Samsung Electronics, SK Hynix, and SK Telecom.

SK Telecom had invested $100 million in Anthropic in 2023 and formed a commercial partnership to develop telecommunications-focused AI models — by any normal measure, a flagship strategic partner, not a security risk. That changed within days. Wired identified SK Telecom as the South Korean telecom company whose access to Claude Mythos the White House ordered revoked over alleged ties to China — access was revoked days before the administration imposed export controls that pulled Anthropic's most advanced AI models offline entirely for all foreign nationals.

A Suspicion, Not a Finding

What makes this story remarkable is the thinness of the evidence behind it. SK Telecom's footprint in China is minimal, generating roughly $1.9 million in Chinese revenue in 2024 and employing seven people in the country, according to its annual report. US officials raised concerns about SK Telecom's historical business ties to China, despite the company's limited current operations there.

SK Telecom has firmly rejected the allegations. SK Telecom has denied the allegations, telling a Korean newspaper that the anonymous claims in foreign media "lack verified facts," and that it has no ties to China. Notably, KT and LG Uplus — the other major Korean carriers — explicitly stated they did not receive access to the Mythos programme, while reports noted that LG U+, the only Korean operator using Huawei equipment in its 5G network, also denied any connection — underscoring just how indirect and unverified the China-link suspicion against SK Telecom appears to be.

From One Revocation to a Global Shutdown

The consequences escalated far beyond a single company's access credentials. The White House later asked Anthropic to revoke SK Telecom's access amid concerns about alleged China ties, and Anthropic immediately complied — though there was no mention of the telecom in the government's subsequent formal demand to restrict Mythos and Fable 5 to US nationals only.
The situation escalated further when Amazon researchers identified potential vulnerabilities in Fable 5, a safeguarded version of Mythos released to the public. The administration ultimately ordered Anthropic to restrict access to both models to US nationals only. Rather than implement nationality-based access controls, Anthropic chose to disable the models entirely — a decision that took two of the world's most advanced AI systems offline globally, affecting legitimate users in over fifteen countries, because of unverified suspicions about one telecommunications company's historical China ties.

The episode badly damaged the Trump administration's confidence in Anthropic's ability to safeguard sensitive technology — and, as watchchina.ai readers know well, sits squarely within a broader pattern: Washington's national security apparatus is now treating any plausible China connection, however indirect, as sufficient grounds for the most severe restrictive action available.

Sources: Wired, The Washington Post, Slashdot, The Elec, BigGo Finance, Korea JoongAng Daily, LLMBase