Anthropic suspends two models under US directive

Anthropic says it has disabled access to Claude Fable 5 and Claude Mythos 5 for all customers after receiving, on June 12, 2026 at 5:21 p.m. ET, a US government export control directive. According to the company, the directive requires suspension of access to the two models by any foreign national, whether inside or outside the United States, including Anthropic’s own foreign national employees. To comply quickly, Anthropic says it shut off both models for its entire customer base, while stressing that access to all other Claude models is not affected.

The notable point is not only the outage, but the reasoning behind it. Anthropic says the government letter did not provide specific details about the national security concern. The company says its understanding is that the government believes it became aware of a method for bypassing, or “jailbreaking,” Fable 5. A jailbreak is an attempt to make a model ignore its safety rules. Anthropic says it reviewed a demonstration of the technique and believes Fable 5 refused the prohibited requests in the tests it was able to run, but it is applying the directive regardless.

The episode gives a concrete shape to a changing AI policy debate. Until now, restrictions have mostly centered on chips, compute infrastructure, sensitive data, or access to particular markets. In this case, based on Anthropic’s account, the measure targets access to named AI models, and the constraint reaches internal staff according to nationality. That creates a difficult operating triangle: obey a public order, keep enterprise service predictable, and preserve researchers’ ability to test model safety under controlled conditions.

For users, the immediate change is straightforward: Fable 5 and Mythos 5 are unavailable through Anthropic, while other Claude models remain available. For the industry, the bigger signal is operational. National security rules can now translate into a fast software cutoff, not only into lists of components, suppliers, or customers. AI providers will need more precise records of who can access which models, in which jurisdictions, and under what shutdown procedures when regulation moves faster than normal product cycles.