White House Forced Anthropic to Cut SK Telecom’s Access to Claude Mythos

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. The directive was based on alleged ties between the South Korean telecom giant and China, according to reporting by WIRED. The incident places SK Telecom at the center of a growing US government push to control who can access frontier AI systems, and signals that export control enforcement is now reaching into commercial AI contracts.

What happened

The White House told Anthropic to cut off SK Telecom’s access to Claude Mythos, the company’s most advanced set of AI models at the time. The order arrived just days before Anthropic took those models offline entirely, according to WIRED. The basis for the directive was alleged ties between SK Telecom and China, though the specific nature of those ties has not been fully detailed in public reporting.

SK Telecom is South Korea’s largest telecommunications company. Its involvement with Anthropic placed it at the center of a US government review that ultimately ended in a forced revocation of access, a notable step in how Washington is applying export control logic to AI model access.

Why it matters

This case is significant for a few reasons. First, it shows the US government is now willing to directly intervene in the commercial relationships of AI companies, not just restrict chip exports or data center buildouts. Second, it sets a precedent: any company with perceived connections to China could find its access to frontier AI models revoked, regardless of where it is headquartered.

For businesses that rely on API access to frontier AI models, this introduces a new category of risk. Access can be cut off not because of anything a customer did directly, but because of political pressure applied to the AI provider. That is a meaningful operational dependency most companies are not currently accounting for.

The timing is also notable. The White House acted just before Anthropic pulled the models offline anyway. That sequence raises questions about whether the two events were connected, or whether the government moved quickly knowing a window was closing.

Our take

From where we sit, this story is less about SK Telecom specifically and more about what it reveals for anyone building on top of a third-party AI provider.

Most businesses treat API access to models like Claude or GPT as a utility, something that is always on unless you stop paying. This case is a reminder that access to frontier AI is closer to a licensed product, one that can be revoked by the vendor under government pressure, terms-of-service changes, or model deprecation.

A few things worth watching:

• Whether other non-US companies receive similar access revocations based on alleged China ties.
• How Anthropic and other frontier AI labs update their terms of service to reflect government compliance obligations.
• Whether export control frameworks are formally extended to cover AI model API access, not just hardware and software traditionally covered under EAR or ITAR.

The honest read here is that the US government is improvising. There is no clean legal framework for AI export controls yet. What exists is political will, and companies like Anthropic are being asked to act on it. That makes the environment unpredictable for any international customer of a US-based AI lab.

What to do about it

If your business or your clients depend heavily on a single AI provider’s API, now is a reasonable time to map that dependency. Consider whether a fallback model or provider is technically feasible, and check whether your vendor’s terms of service include language around government-directed access revocation. You do not need to act today, but you should know where you stand before access disappears without warning.

Source: WIRED · AI