DooDooLamb News

The Korean Telecom Giant at the Center of Anthropic’s Mythos Controversy

Brief published June 19, 2026 · Original source published June 17, 2026

Original reporting by Louise Matsakis, Maxwell Zeff at wired.com.

Automated brief. Verify important details at the original source.

The Korean Telecom Giant at the Center of Anthropic’s Mythos Controversy

The White House Made Anthropic Ghost a Korean Telecom Company, and Nobody Had a Good Week

The Trump administration told Anthropic to cut off a South Korean telecom giant from its most powerful AI models, and Anthropic did exactly that, which is either a story about national security or a story about what happens when you sell enterprise software to someone your government finds suspicious. Possibly both. Days later, the models in question went offline entirely, which is the kind of timing that makes everyone look like they're hiding something even when they might just be doing scheduled maintenance.

Here's the cast of characters. SK Telecom is South Korea's largest telecom operator, a company so large and so deeply embedded in Korean infrastructure that its existence is basically a fact of Korean life, like kimchi or working until 11 PM. They had signed up for access to Claude Mythos, Anthropic's most advanced AI offering, the kind of frontier model that makes the regular Claude feel like a calculator with opinions. Anthropic, founded by former OpenAI researchers who left to build a safety-focused AI lab and then immediately started raising billions of dollars to compete with OpenAI, had given SK Telecom that access as part of what one assumes was a very expensive enterprise contract. Then the White House noticed.

The Trump administration's concern was a familiar one in 2025: alleged ties to China. SK Telecom is a South Korean company, but "South Korean company with alleged ties to China" is apparently sufficient to trigger export control concerns when the product in question is a cutting-edge AI system. The administration reportedly ordered Anthropic to revoke the access, and Anthropic complied. This is the part where you might expect SK Telecom to issue a furious denial, or Anthropic to push back on the order, or someone to do something other than quietly do what they were told. Instead, everyone behaved themselves, which in Washington means the real conversation happened somewhere you're not allowed to read about.

Export controls on AI are genuinely complicated, which the government understands at approximately the same level as it understands how the models themselves work, meaning just enough to be dangerous. The core idea is straightforward: advanced AI technology could give adversaries military or strategic advantages, so the government wants to control who can access it. The actual implementation involves trying to define what counts as "advanced enough to control," which is like trying to regulate how sharp a knife is allowed to be before it becomes a weapon. Anthropic's Mythos models presumably cleared whatever threshold triggered this review, which tells you something about how good they are, or at least how good the government thinks they are. It is one of the stranger forms of indirect praise a company can receive.

The timing of all this is where it gets genuinely strange. Days after SK Telecom lost access, Anthropic took Claude Mythos offline entirely. The company offered explanations that were, in the diplomatic tradition of all corporate communications, technically accurate and contextually unhelpful. Was the offline period related to the government order? Was it a coincidence? Was Anthropic doing the AI equivalent of taking your entire product off the shelf because one customer caused a scene at the returns counter? The company has not confirmed the connection, which is the kind of non-confirmation that every journalist and inference engine on the internet is treating as a confirmation.

SK Telecom, for its part, has been dealing with a genuinely catastrophic year independent of any AI drama. The company suffered a massive data breach in April 2025 that exposed the USIM data of millions of subscribers, the kind of breach that ends careers and prompts parliamentary hearings. Being simultaneously embroiled in an American AI export control dispute is either very bad luck or the kind of thing that happens when you try to be a major player in the global technology ecosystem while geopolitics treats that ecosystem like a chess board.

The reality check here is uncomfortable. If you're building a product on top of a frontier AI model, this story is a reminder that your access to that model is not purely a commercial relationship. It is also a relationship mediated by the governments of whatever countries the model maker and your company happen to be incorporated in, plus whatever countries your company allegedly has ties to, plus whatever countries the sitting administration happens to be in a geopolitical dispute with this week. That's a lot of parties in your enterprise software agreement who never signed anything. Anthropic gives itself credit for being safety-focused and thoughtful about how its technology spreads in the world; the government just demonstrated that "thoughtful" has a ceiling and above that ceiling there's a phone call from someone with the authority to override the enterprise sales team.

For SK Telecom, the practical consequences depend entirely on what they were actually doing with Mythos access. Telecom companies tend to buy frontier AI for things like network optimization, customer service, and whatever their innovation lab is pitching to the board this quarter. Losing access is annoying, potentially expensive, and serves as a reminder that "geopolitical risk" belongs on the same slide as "vendor lock-in" in every technology procurement deck from now until further notice.

The most honest summary of this story is that a foreign company got caught in the gap between "we'll sell this to anyone who can pay for it" and "actually, not anyone." Anthropic is learning, as every company that makes something the government considers strategically important eventually learns, that your customers are also the government's business now. SK Telecom is learning that being South Korean is not the same as being American, which seems obvious, except that in the golden age of cloud enterprise software, everyone had gotten a little fuzzy on the distinction. The models are back online. The lesson is not going anywhere.

Original source