Anthropic's Fable 5 and Mythos 5 Models Taken Offline After US Export Control Order Sparks Global Backlash

Anthropic's two most capable AI models are no longer available to any customers anywhere in the world, after the US government issued an export control directive that forced the company to pull them entirely rather than attempt to enforce the restrictions on a user-by-user basis. The affected models — Fable 5 and Mythos 5 — were disabled across all of Anthropic's consumer and enterprise products, including its Claude.ai interface and API, almost immediately after the government order was issued.

The directive, which reportedly cited national security grounds, bars foreign nationals from accessing the models regardless of whether they are located inside or outside the United States. Because Anthropic's user base is international and because reliably verifying the nationality of every user in real time would be technically complex, the company chose to suspend access universally until it can establish a compliant path forward. Anthropic acknowledged the shutdown in public communications, noting it had disabled the models to ensure full compliance with the order.

The move drew a swift and coordinated response from the cybersecurity and AI research communities. More than sixty security professionals, and subsequently over one hundred signatories, put their names to an open letter calling on the Trump administration to reverse the directive. The letter argued that restricting access to frontier AI models harms defensive security practitioners — researchers, threat analysts, and incident responders — far more than it constrains the adversaries the policy is ostensibly designed to contain.

Critics of the export order pointed to the nature of the research that reportedly triggered it. According to security expert Katie Moussouris, who reviewed the underlying third-party study, researchers had used entirely routine security prompts — asking the model to review and fix vulnerable code — rather than exploiting any novel loophole or bypass. There was no jailbreak, in her assessment. The model's behavior was consistent with standard defensive security workflows, and restricting it on that basis sets a precedent that could chill legitimate security research conducted with AI assistance.

The geopolitical backdrop adds complexity to the debate. Anthropic has itself voiced concern about the practice of model distillation, through which rival developers extract capabilities from American AI systems to train competing models. Chinese AI companies, including DeepSeek, have been accused by Anthropic and Google of using this technique. However, critics of the export control note that it cannot easily extend to open-weight models or to AI systems already deployed outside US jurisdiction, making the ban's practical effect on foreign adversaries limited while its cost to US-aligned defenders is concrete and immediate.

The timing of the shutdown also intersected with Anthropic's separate decision to update its consumer privacy policy. That revised policy, taking effect July 8, 2026, grants the company authority to conduct identity and age verification checks on individual subscribers — a change that may in part be designed to give Anthropic better mechanisms for enforcing export compliance in the future without having to resort to blanket shutdowns.

As of publication, neither Fable 5 nor Mythos 5 had been restored. Anthropic said it was working to understand the scope of the directive and determine what steps would be needed to bring the models back online in a manner consistent with both legal requirements and its obligations to customers.