The US government issued an export control directive ordering Anthropic to suspend all access to Claude Fable 5 and Claude Mythos 5 for foreign nationals, citing national security authorities without elaboration. Anthropic responded by blocking public access to both models entirely, globally. This means no users worldwide can currently access them, including paying enterprise customers and internal employees.

The move represents an abrupt reversal. Anthropic had recently released Fable 5 and Mythos 5 as its most powerful generally available models, positioning them for widespread commercial adoption. The government order transformed a public product launch into a complete shutdown within hours.

The directive signals a toughening stance on AI model exports. Rather than implementing granular geographic restrictions, the government forced a blanket prohibition. Anthropic chose the nuclear option of removing access entirely rather than maintaining a fragmented system where some users could access the models and others couldn't.

For enterprises relying on these models, the implications are immediate and severe. Organizations that integrated Fable 5 or Mythos 5 into production systems now face service interruption with no clear timeline for restoration. The company has not specified when or whether these models will return to public availability, or what conditions might allow their release.

This precedent matters. The government's willingness to invoke export controls on AI models after public release sets a new baseline for regulatory risk. Other AI companies may face similar directives. The action also highlights that national security determinations around AI happen behind closed doors, with companies and users learning only after restrictions take effect.

Anthropic published a statement acknowledging the government order but provided limited detail on affected users or timeline for remediation. The company urged affected parties to reach out directly, but enterprise customers remain in limbo regarding alternatives or compensation.

The incident demonstrates that regulatory intervention in AI deployment can happen rapidly and comprehensively, bypassing grad