Anthropic Under Fire: Allegations of Chinese Ties and the SK Telecom Case

What happened?

According to an investigation by Wired published in March 2025, early that year the White House ordered Anthropic to revoke access to its most advanced model, internally known as 'Claude Mythos', from South Korean telecom giant SK Telecom. The official reason: alleged ties between SK Telecom and China that could violate U.S. technology export control rules. Days later, Anthropic removed access to Mythos for all clients, not just SK Telecom, sparking speculation about whether the move was preventive or government-mandated. Internal sources cited by Wired indicate the order came directly from the Executive Office of the President, though Anthropic has publicly stated the decision was its own. The case is part of the growing U.S.-China rivalry over AI dominance and marks a milestone at the intersection of frontier technology and geopolitics.

Why is it important?

This case is the first where a frontier AI model has been withdrawn for geopolitical, not technical, reasons. Until now, models were disconnected due to security flaws or biases, but never due to external political pressure. This sets a precedent: AI companies will have to comply with government directives that go beyond technical safety, delving into national security. Moreover, it reveals that Anthropic, despite its rhetoric of transparency and responsibility, can act under political pressure without detailed explanations. The controversy also casts doubt on the reliability of AI model risk assessments, as Mythos was rated 'safe' before being disconnected. Cybersecurity experts note this could erode trust in model certification processes, which are currently voluntary in the U.S. unlike the EU with the AI Act.

Historical context is key: since 2022, the U.S. has tightened export controls on semiconductors and AI technology to China, but this is the first time it has been applied to an AI model itself, not hardware. Compared to the 2022 blockade of Nvidia chips to China, here the object of control is intangible: access to a language model. This opens a new front in the tech war, where AI models become strategic assets. For users and businesses, it means access to the most advanced AI may be subject to political restrictions, creating uncertainty in long-term technology planning.

Consequences for the sector

• Export controls: Other AI companies, such as OpenAI and Google, are expected to face similar pressures to vet foreign clients. In fact, according to Wired sources, the Biden administration had already contacted OpenAI in 2024 about Chinese clients, but without issuing a direct order. This case could be the first in a series.
• Trust in transparency: Anthropic's lack of clear communication may erode trust among developers and clients. If a company that promotes 'responsible AI' hides government pressure, what can be expected from others? This could lead to increased demand for independent audits and transparency in access decisions.
• Market fragmentation: Non-U.S. companies may seek local or Chinese alternatives to avoid geopolitical dependency. For example, SK Telecom had already invested in Anthropic in 2023 and is now accelerating its own AI model with local partners. This could accelerate the creation of regional AI ecosystems, such as in Europe with Mistral or in China with Baidu and Alibaba.
• Impact on startups: Startups relying on Anthropic's API for their products now face the risk of their clients being vetoed for political reasons. This adds a layer of geopolitical risk to AI startup evaluation, similar to that already existing in sectors like defense or energy.

In market terms, the withdrawal of Mythos could slow the adoption of frontier models in sensitive sectors like telecommunications or finance, where geopolitical stability is key. According to Gartner analysts, the global AI market could segment into three blocs: the U.S., China, and the rest, with increasingly high technical and political barriers.

What should readers know?

Anthropic has stated the decision was its own and not due to government order, but internal sources cited by Wired indicate otherwise. The company claims the global withdrawal was for 'internal security reasons' but has not provided further details. SK Telecom, for its part, denies any illegal ties with China and says it will comply with regulations. The story continues to unfold: congressional hearings and potential sanctions are expected. Additionally, the U.S. Department of Commerce is reviewing whether current export control rules cover AI models, which could lead to new specific regulations. For readers, it is crucial to understand that frontier AI is no longer just a technical matter but a geopolitical battlefield. Trust in AI companies' transparency is at stake, and market fragmentation could have long-term consequences for innovation and global access to technology.

“This is just the beginning. AI has become a geopolitical battlefield,” said an analyst consulted by TheVortiq. “What we are seeing is the start of an era where AI models are treated as weapons or strategic resources, with all that implies for technological sovereignty and international cooperation.”