White House Ordered Anthropic to Cut SK Telecom From AI Model Access

Days before Anthropic took its most advanced AI models offline, the White House ordered the company to revoke SK Telecom’s access to Claude Mythos over claims of alleged ties to China. The intervention, reported by Wired, marks one of the first direct executive branch actions targeting a specific foreign entity’s access to frontier AI capabilities. The incident underscores a growing tension between global AI development and national security concerns, with implications that reach far beyond a single telecom company.

The Incident and Its Immediate Fallout

Anthropic, the company behind the Claude family of AI models, had granted SK Telecom access to its most powerful system, codenamed Mythos. The South Korean telecom giant had been a strategic partner, exploring applications of advanced AI in telecommunications infrastructure. But the White House moved swiftly, citing concerns that SK Telecom’s corporate structure and partnerships could create channels for technology transfer to China. Anthropic complied with the order, revoking access and subsequently taking the Mythos models offline entirely. The company has not publicly detailed the technical or contractual steps involved, but the speed of the action suggests that the White House invoked export control authorities that treat access to frontier AI models as a controlled technology transfer.

Export Controls Enter the AI Era

This event signals a paradigm shift in how governments view advanced AI models. Previously, export controls focused on hardware like semiconductors and specialized chips. Now, access to the models themselves, even through cloud APIs, is becoming a regulated activity. The SK Telecom case establishes a precedent: any foreign entity with perceived national security risks, including those with complex supply chains or ownership structures, could face similar restrictions. For AI companies, this means that partnerships with foreign telecoms, banks, or energy firms may now trigger compliance reviews. The legal basis for these controls remains murky, as existing export control laws were not written with AI models in mind. This creates uncertainty for both AI developers and their enterprise customers.

Implications for Practitioners and Decision Makers

For AI practitioners, the main takeaway is that model access can be revoked without notice due to geopolitical factors. Companies building applications on top of frontier models from providers like Anthropic must now consider geopolitical risk as part of their technology stack. They should diversify model providers and build fallback mechanisms that do not rely on a single API. For decision makers at AI companies, the lesson is clear: compliance teams must expand their scope beyond data privacy to include export control law. The White House’s willingness to intervene directly suggests that future partnerships with foreign entities, especially in telecommunications and energy, will face heightened scrutiny. The incident also raises questions about the liability of AI companies if their models are used by sanctioned entities through indirect access.

What to Watch Next

The SK Telecom case is likely the opening move in a larger regulatory push. Expect the Biden administration or its successor to formalize rules around AI model access, possibly through the Commerce Department’s Bureau of Industry and Security. Companies like Anthropic, OpenAI, and Google will face pressure to implement real-time geolocation and identity verification for API access. For the broader industry, the key question is whether these controls will be applied uniformly or create a patchwork of restrictions that stifle innovation. The next few months will reveal whether the White House views this action as a one-off or the beginning of a sustained campaign to control the global flow of AI capabilities.