The US government issued a directive on June 12 ordering Anthropic to immediately suspend access to its most advanced AI models for all foreign nationals. This mandate includes the company’s own international employees and targets the newly released Fable 5 and Mythos 5 systems.
A reported jailbreak vulnerability triggered national security alarms, raising concerns that adversaries could exploit the models to identify software weaknesses. Anthropic complied swiftly, disabling global access to both systems. This marks the first time US export controls have compelled a major AI lab to fully revoke access to frontier models rather than issuing fines or warnings.
The directive originated from the Trump administration as part of a broader strategy treating advanced AI technologies as national security assets. While Anthropic implemented the ban immediately, citing the infeasibility of targeted restrictions on short notice, the company publicly criticized the approach. Leadership argued that a potential vulnerability does not justify halting international access to commercially deployed products and called for a more transparent regulatory process.
Cybersecurity experts warn that jailbreaks bypass safety guardrails, potentially turning these models into tools for offensive cyber operations. However, over a dozen experts have also criticized the blanket ban as a dangerous move that stifles innovation without guaranteeing security.
For investors, this episode introduces significant regulatory risk to the AI sector. The precedent establishes that commercially deployed AI products can face immediate international shutdowns based on security concerns. Enterprise clients integrating these models suddenly lost access, signaling that any frontier lab, including OpenAI and Google DeepMind, now faces similar exposure to ad hoc government intervention.