Those concerns were heightened when Amazon later flagged vulnerabilities in the White House that it identified in Fable 5, a highly secure version of Mythos that Anthropic released to the public on June 9. Amazon researchers claimed that it was possible to bypass some of Fable 5’s guardrails and access Mythos’ formidable cyber capabilities, although Anthropic and outside cybersecurity experts have argued that these risks are not unique to the cloud.
According to a person close to the administration, the confluence of events ultimately led the White House to decide that it could no longer trust Anthropic to protect its most advanced AI technology. On Friday, the Trump administration ordered Anthropic to revoke access to Mythos and Fable 5 for all foreign nationals, including immigrants, inside the US.
Rather than restrict access to its technology based on nationality, a process that would be difficult to implement while maintaining privacy, Anthropic decided it would be better to disable access to the models altogether. The White House and Anthropic remain at odds after days of talks about the cloud mythos and bringing Fable 5 back online.
Anthropic declined to comment. The White House and SK Telecom did not immediately respond to requests for comment. The Washington Post previously reported that Trump administration officials became concerned after learning that Mythos recipients included a “South Korean telecommunications company” they believed had ties to China, although the article did not name the company. In response to that reporting, SK Telecom told a Korean newspaper that “the anonymous insider’s comments in foreign media lack verified facts, and our company has no ties to China.”
A person close to Anthropic said the company views SK Telecom’s access to Mythos and the vulnerabilities identified by Amazon as separate issues. He noted that the letter the US government sent to Anthropic, demanding that it restrict access to Cloud Mythos and Fable 5 to US citizens only, did not reference the Korean company or China.
Because CloudMythos is exceptionally adept at identifying software vulnerabilities, Anthropic restricted early access to a small group of trusted organizations through a program called Project Glasswing. Earlier this month, SK Telecom, South Korea’s largest wireless carrier, became one of about 150 companies to gain access to Mythos as Anthropic expanded the program “after several weeks of close collaboration” with external experts and the U.S. government.
SK Telecom has injected capital into Anthropic several times, including a $100 million investment in 2023, which coincides with the formation of a commercial partnership to develop AI models tailored to the telecommunications industry. It was one of several Korean organizations participating in Project Glasswing, along with Samsung Electronics and the Korea Internet and Security Agency.
Earlier this month, shortly after Anthropic announced the latest expansion of Project Glasswing, the White House asked Anthropic to revoke SK Telecom’s access to Mythos, according to a person associated with the AI lab. The company immediately complied and the U.S. government did not threaten to impose export controls on the model at the time, sources told WIRED.
While SK Telecom does not appear to have major operations in China, it is part of a much larger conglomerate called SK Group, whose affiliates maintain extensive business interests in semiconductor, energy and other industries in the country.
<a href