US directs Anthropic to restrict Claude Fable 5 and Mythos 5

The U.S. Commerce Department issued an export control directive on June 12 that targets Anthropic’s newest AI models, and Anthropic responded by suspending access globally. The move sharpens the fight over frontier AI security, model access, and how far export controls should reach. Why it matters: - The directive puts U.S. national security concerns directly into the rollout of frontier AI models. - The global suspension shows how hard it can be for AI companies to enforce nationality-based access rules in real time. - The move adds pressure on regulators and AI developers to set clearer standards for safety testing, export controls, and deployment oversight. What happened: - The U.S. Department of Commerce issued an export control directive on June 12 requiring Anthropic to restrict foreign nationals’ access to Claude Fable 5 and Mythos 5. - Anthropic suspended access to both models worldwide after determining it could not reliably verify users’ nationalities in real time. - A U.S. government official confirmed that the Commerce Department sent the directive to Anthropic. - Other Claude models remain available to users. The details: - Claude Fable 5 and Mythos 5 launched on June 9 as Anthropic’s latest frontier AI models. - Mythos 5 had previously been limited to vetted organizations through Anthropic’s Project Glasswing initiative. - Fable 5 was designed for broader deployment with advanced capabilities and additional safety measures. - The action reportedly followed the discovery of a technique that could let users bypass certain safeguards built into the models. - Anthropic said it disagrees with the characterization of the vulnerability as a significant threat. - Anthropic said similar issues could be identified using other publicly available AI systems. Between the lines: - The directive reflects growing government attention to the security risks tied to the most capable AI systems. - The episode highlights a practical enforcement problem: access controls tied to nationality are difficult to apply cleanly on global AI platforms. - The dispute also shows a wider gap between regulators focused on model risk and developers who may view certain vulnerabilities as manageable or non-unique. What’s next: - Anthropic and federal authorities are expected to keep talking about the deployment and governance of advanced AI technologies. - Policymakers, researchers, and industry stakeholders are likely to keep pushing for clearer rules on AI safety testing and export controls. - The case may become a reference point in future debates over whether frontier models should face tighter access restrictions. The bottom line: - The Commerce Department’s move and Anthropic’s global suspension mark a sharp escalation in the fight over who can access the most advanced AI models.