Anthropic Pulls Fable, Mythos After Government Issues Emergency Export Control Order

The U.S. government forced Anthropic to halt distribution of its Fable 5 and Mythos 5 AI models—just days after CEO Dario Amodei explicitly called for such regulatory oversight. On June 13, 2026, the Department of Commerce issued an emergency export control order, citing national security risks posed by the advanced models' potential misuse in adversarial hands. Anthropic complied immediately, pulling the models from its cloud platforms and halting downloads. The move marks an unprecedented escalation in AI governance: for the first time, a sitting government has used emergency powers to freeze deployment of a frontier AI system before a vulnerability was publicly disclosed. Anthropic, a company founded on principles of responsible AI development, had voluntarily released Fable 5 and Mythos 5 only weeks earlier after extensive safety testing. But the government's decision suggests that even rigorous internal evaluations were insufficient to address intelligence community concerns about dual-use capabilities—particularly in areas like cyber offense, synthetic biology, and disinformation at scale. Dario Amodei, co-founder and CEO, had published an op-ed just days prior in which he urged governments to adopt 'predictable, preemptive oversight' for advanced AI models, warning that the window for safe deployment was closing. His call now appears prescient, as the order he advocated for arrived almost immediately. The emergency export control order leverages the International Emergency Economic Powers Act (IEEPA) and imposes a 40-day freeze on any transfer of Fable 5, Mythos 5, or derivative technologies to foreign entities. Companies violating the order face penalties of up to $1 million per incident. Anthropic stated it would cooperate fully and even offered to share its safety evaluations with the government. However, the abruptness raised concerns among investors and developers who had built applications relying on these models. Analysts point to two implications. First, Washington is signaling that AI export controls—previously limited to semiconductors—now extend to software and model weights. Second, the U.S. is drawing a line that other allied nations may follow, potentially fracturing the global AI supply chain. The order also pressures other frontier labs like OpenAI and Google DeepMind to preemptively engage with regulators. 'This is the beginning of a new era in AI governance,' said Dr. Anika Patel, a fellow at the Center for AI and Digital Policy. 'We are moving from voluntary commitments to enforceable mandates.' The next 40 days will be critical: the government will review Anthropic's safety data and decide whether to extend the freeze permanently, request modifications to the models, or lift the order. Meanwhile, competing AI firms are watching closely—and quietly assessing their own exposure to similar orders. For Anthropic, it's a stress test of its founding mission: to build AI that is safe, transparent, and aligned with human interests. The irony is that its CEO's plea for oversight may have made the company the first test case of that very principle.