US orders Anthropic to pull newest AI models over national security fears

Anthropic announced it has taken its latest artificial intelligence models offline following a strict federal directive.

The San Francisco-based firm disabled access to Fable 5 and Mythos 5 after receiving an export control order from the United States government on Friday, June 12.

Citing national security authorities, the legal directive restricts any foreign national, whether inside or outside the country, from accessing these specific systems.

Anthropic explained that the mandate covers its own foreign employees, making a complete, abrupt shutdown for all commercial customers the only viable way to comply.

Access to the standard Claude chatbot and other older models remains entirely unaffected by the sudden regulatory intervention.

According to the developer, the federal letter arrived late in the afternoon and lacked granular technical details regarding the specific national security threats.

The company understands that officials became alarmed by a potential method to bypass, or jailbreak, the safety guardrails integrated into Fable 5.

Anthropic strongly corporate-disagrees with the sweeping enforcement action, labeling the incident a misunderstanding, and it is actively lobbying to restore normal service.

Engineers reviewed the technical report believed to have triggered the intervention, asserting the flagged vulnerability is minor, narrow, and non-universal.

The specific technique involves prompting the system to review a codebase and repair software flaws, a capability Anthropic claims is standard across competing systems.

Company officials stated that the demonstrated capability level is already widely available in other models, such as OpenAI's GPT-5.5, and is regularly used by cybersecurity defenders.

Prior to the commercial release of Fable 5 on June 9, the firm conducted extensive red-teaming evaluations spanning thousands of hours.

These safety trials involved collaboration with the United Kingdom Artificial Intelligence Safety Institute (UK AISI), domestic government bodies, and external specialists.

The developer defended its defense-in-depth safety strategy, which relies on strict monitoring and a mandatory 30-day customer data retention policy to track down malicious exploitation.

Company executives warned that pulling a commercial product deployed to millions over a narrow, non-universal vulnerability sets a dangerous precedent that could halt industry-wide innovation.