Anthropic suspends Claude Fable 5 access following US government directive

Anthropic has taken its Claude Fable 5 and Mythos 5 artificial intelligence models offline to comply with an immediate export control directive issued by the United States government. The order, received on Friday at 5:21pm ET, cites national security concerns and specifically targets access for foreign nationals, whether inside or outside the United States. To ensure full compliance with the directive, Anthropic has suspended access for all customers, not just foreign nationals.

The company stated that the government’s belief that a method to 'jailbreak' Fable 5 had been identified triggered the order. In a blog post, Anthropic noted that the letter from US authorities did not provide specific details of the national security concern. The firm explained that its understanding is that the government believes it has become aware of a technique to bypass the model's safeguards.

Anthropic reviewed a demonstration of the alleged jailbreak technique, which involved asking the model to read a specific codebase and fix software flaws. The company argued that the identified vulnerabilities are 'relatively simple' and that other publicly available models can discover them without requiring a bypass. Anthropic contends that the jailbreak is narrow and non-universal, meaning it does not make an attacker meaningfully more dangerous than they would be with other AI models.

This incident marks a further escalation in tensions between Anthropic and the Trump administration. Earlier this year, the Department of Defense labeled Anthropic a 'supply chain risk' after the company drew red lines over how the US military could use its technology. That designation effectively barred government agencies and contractors from using Anthropic’s technology, prompting Anthropic to file lawsuits against the administration.

Claude Fable 5 was publicly released on Tuesday, a version of the Mythos AI model with safeguards preventing answers about cybersecurity, biology, and chemistry. The public release was conducted in collaboration with the US government, following a limited rollout of the Mythos Preview AI model in April aimed at helping organisations improve their cybersecurity defences. Spokespeople for the White House and US Commerce Department did not immediately respond to requests for comment.

Anthropic CEO Dario Amodei had previously published a policy essay supporting a structured government process for blocking unsafe AI models. However, in its statement regarding the current directive, the company argued that this specific action does not adhere to those principles of fairness and transparency. The suspension affects all users globally, regardless of location, as the firm prioritised compliance with the US government's order.