Anthropic blocked by White House: Fable 5 and Mythos 5 shutdown

Anthropic has taken its newest models, Fable 5 and Mythos 5, offline after a June 12 directive from the White House that ordered the company to block access for foreign nationals. The company said it complied with the government directive and removed access to both models for all customers.

What happened

Anthropic launched Claude Fable 5 on June 9, calling it "the most powerful model it has ever made widely available" and saying "Fable 5’s capabilities exceed those of any model we’ve ever made generally available." Mythos 5 shares the same underlying model, with Anthropic saying Mythos had "the safeguards lifted in some areas."

On June 12 the government directed Anthropic to block access to Fable 5 and Mythos 5 for foreign nationals, a move the company said it could not implement by excluding only non-U.S. users and therefore cut access to the models for all customers. Anthropic stated, "We are complying with the government’s legal directive and are removing access to Fable 5 and Mythos 5 for all users." The company also said the government did not provide specific details of its national security concern and that the evidence of a potential jailbreak was provided verbally.

Coverage since the order has added detail and differing accounts. The Wall Street Journal reported the export control directive was triggered in part by cybersecurity research from Amazon after Amazon presented findings to the White House. That research reportedly showed a series of prompts that could get Fable 5 to provide information that could be used in cyberattacks. Semafor reported the decision was also driven in part by fears that a group linked to China may have accessed Mythos.

Before the export directive, Anthropic was already involved in a separate dispute with the government over the Pentagon. After the June 12 instruction prevented foreign access — including for foreign-born Anthropic employees — the company chose to cut access entirely rather than segregate users.

What Anthropic and others are saying

Anthropic pushed back on the scale of the concern. The company said the government shared only a narrow finding about a potential jailbreak and that those vulnerabilities were minor and available in other models, naming GPT 5.5 as an example. Anthropic said it disagreed that a narrow potential jailbreak should require recalling a commercial model deployed widely.

The White House has not publicly confirmed all of the published reporting. Semafor noted the White House did not confirm a China link, and a spokesperson to Semafor said China was not raised during government discussions around export controls. A post on X by a Trump adviser focused on reported jailbreak ability rather than foreign access. Amazon has not publicly commented on the report that its research prompted the White House conversations.

Why it matters

The order shows the U.S. government is willing to use export controls and access restrictions to limit who can use frontier AI models. That approach complicates product deployment for companies with international users and multinational workforces, and it raises questions about how firms will balance government directives, researcher access, and product availability. It also highlights how private security research can quickly shape government action on AI if officials conclude a model poses a risk.

What to watch

Watch for formal statements or evidence from the White House clarifying the national security concern and any technical details of the alleged jailbreaks. Also watch whether Amazon or other firms publish their security findings, and whether Anthropic re-releases gated versions of the models or alters safeguards to address the government's concerns.