Anthropic forced to suspend new AI models after U.S. export-control order

Anthropic said it was forced to disable its newly released AI model Fable 5, along with the more advanced Mythos 5, after receiving an order from U.S. authorities. The company said the practical effect of the directive was to stop access to both models for all customers so it could comply with the rules. The order, issued on June 13, 2026, cites national security authorities and bars any foreign national, including Anthropic employees who are not U.S. citizens, from accessing the systems inside or outside the United States.

Fable 5 is the public, safety-limited version of Mythos 5, which Anthropic says has no safety barriers and is restricted for now to a small group of defensive cybersecurity firms and critical infrastructure providers working with the U.S. government. Anthropic has described Mythos 5 as unusually powerful and dangerous, saying it is “too powerful” compared with any previous public model. The company said that is why it initially released Mythos 5 only to a limited set of security and defense organizations in the United States.

Anthropic said U.S. national security officials did not point to a specific flaw in Mythos or Fable. Instead, the company said the government believes it found a way to bypass Fable 5’s protections and perform a jailbreak, meaning to circumvent safety limits and potentially expose sensitive information or enable prohibited uses. Anthropic said its own review found only a small number of known, minor weaknesses, and that similar issues exist in other models without undermining their safeguards.

Before launch, Anthropic said it had invested in protections to prevent the models from being used to hack systems or abuse cyber capabilities. Still, government, technology and financial-sector figures had raised concerns about a wide release. In April, a limited early access program for Mythos 5 was given to a small number of organizations for demonstrations and vulnerability testing.

The controversy comes amid an ongoing fight between Anthropic and the Trump administration. Trump has publicly criticized the company, and Defense Secretary Pete Hegseth called Anthropic a “supply chain risk,” a label usually reserved for companies from adversarial countries. Anthropic sued the Pentagon over that decision and is also fighting another case involving efforts to block government agencies from using its AI tools. A federal judge has so far ruled the Pentagon directive cannot be enforced, so agencies and military contractors may continue using Anthropic products until the case is decided.