Anthropic has suspended access to its two most advanced AI models, Claude Fable 5 and Claude Mythos 5, after the US government issued an export control directive on national security grounds. The models had been available to the public for only three days.
What happened
Anthropic launched Fable 5 and Mythos 5 on 9 June 2026, describing Fable 5 as the most capable model it had ever released to the public. It posted strong results across software engineering, knowledge work and vision benchmarks.
On 12 June, at 5:21pm ET, the company received a letter from the US Commerce Department. The directive ordered Anthropic to suspend access to both models for any foreign national, whether inside or outside the United States, including the company’s own foreign national employees.
Because Anthropic could not separate that group from its wider user base in time, it disabled both models for everyone. All other Claude models, including Claude Opus 4.8, remain available.
Why the government acted
The letter did not spell out the specific concern. Anthropic has said its understanding is that the government became aware of a method for bypassing, or jailbreaking, Fable 5.
The company reviewed a demonstration of the technique. It said the method surfaced only a small number of previously known and minor software vulnerabilities, and that other publicly available models could find the same issues.
Anthropic’s response
Anthropic has complied with the order while publicly disagreeing with it. In its statement, the company said it believed the situation was a misunderstanding and that it was working to restore access as soon as possible.
It also noted that it had received only verbal evidence of the jailbreak, with no formal written disclosure, and argued that a narrow finding should not be grounds for recalling a model used by a large customer base.
Why this matters for businesses
This is being reported as the first time US export controls have targeted a deployed commercial AI model itself, rather than chips or training hardware. That sets a notable precedent.
For teams building on AI, the lesson is practical. Model availability can change with little warning for reasons outside your control. If your workflows depend on a single model, sudden suspension can disrupt live operations.
The sensible response is to keep your tooling model-agnostic where possible, monitor provider status pages, and have a fallback model ready. For most users, switching to an alternative such as Claude Opus 4.8 is straightforward, but only if your systems are built to allow it.
