US mandates Anthropic block foreign access to advanced AI models

The United States government has issued an export control directive requiring AI firm Anthropic to block access to its Fable 5 and Mythos 5 models for all foreign nationals. The order, received by the San Francisco-based company at 5:21pm ET on Friday, mandates the immediate suspension of access for foreigners both within and outside the United States, including those currently employed by Anthropic.

Anthropic complied with the directive shortly after its receipt, cutting off access for all affected users at short notice. In a statement published on Friday, the company behind the Claude chatbot expressed disagreement with the measure. It noted that the government letter did not provide detailed specifics regarding the threat, though experts assessed the order was likely triggered by a report concerning a limited capability of the AI to review specific programme code and correct errors.

The scope of the directive has drawn scrutiny for its breadth. The New York Times described the order as "unusually expansive," noting that it could prevent Anthropic’s employees in allied nations, such as Canada and the United Kingdom, from using the models. The Fable 5 model, released earlier in the week, has its cybersecurity and biotechnology capabilities blocked for foreign users, while Mythos 5, the non-public full version, remains restricted to government agencies and selected corporate partners.

Anthropic emphasised that similar capabilities exist in rival models, including OpenAI’s GPT-5.5. The company argued that its safety measures have been extensively tested and questioned the necessity of blocking software used by hundreds of millions of users. The Mythos AI model is noted for its ability to detect software vulnerabilities, some of which have remained undiscovered for decades, a capability that has been used by US authorities to plug security gaps but also raises concerns about potential weaponisation for cyberattacks.

The move follows broader US efforts to vet advanced AI systems for national security risks. Earlier in June, US President Donald Trump signed an executive order requiring federal vetting of the most advanced AI systems. This regulatory environment coincides with Anthropic’s own proposals earlier in the month, where the company suggested that top AI firms coordinate to pause the development of advanced systems, warning of risks to human control as technology improves rapidly.