On June 12, 2026, the U.S. Commerce Department ordered Anthropic to immediately block foreign nationals from accessing its advanced Claude Fable 5 and Mythos 5 models, citing national security concerns. Due to the inability to verify the nationality of all users in real time, Anthropic disabled both models entirely for all users within hours of receiving the directive. The department lifted these restrictions on June 30, 2026, after Anthropic implemented more stringent safety guardrails. These new constraints significantly impaired the models' performance; according to experimental testing, the updated Fable 5 completed only 3 of 12 tasks that were previously routine.

The government issued this mandate under the Export Control Reform Act of 2018, which was originally intended to regulate physical hardware exports like uranium enrichment centrifuges. This marked the first instance of export controls being used to limit access to an AI service. Legal scholars note that the order's legal basis is uncertain, as current law typically distinguishes between physical goods and software interactions; previous guidance from the Commerce Department generally considered remote access to U.S.-hosted software to be outside the scope of traditional export controls. Furthermore, the directive was applied unilaterally without the standard public vetting, inter-agency review, or temporary classification processes—such as the 0Y521 code—that typically accompany federal technology restrictions.

Anthropic did not legally challenge the order, characterizing the event as a misunderstanding and opting to negotiate directly with federal officials. This response aligns with public statements from Anthropic CEO Dario Amodei, who argued shortly before the incident that the government should possess the power to block the deployment of dangerous frontier AI models. Additionally, the 2018 statute limits the authority of federal courts to overturn such administrative decisions, requiring challengers to prove a directive is unconstitutional rather than simply unreasonable. This incident signals a shift toward a model of government control where export privileges for AI are conditioned on secretive access to, and modifications of, the most powerful systems currently in development.

On June 12, 2026, the U.S. Commerce Department ordered Anthropic to immediately block foreign nationals from accessing its advanced Claude Fable 5 and Mythos 5 models, citing national security concerns. Due to the inability to verify the nationality of all users in real time, Anthropic disabled both models entirely for all users within hours of receiving the directive. The department lifted these restrictions on June 30, 2026, after Anthropic implemented more stringent safety guardrails. These new constraints significantly impaired the models' performance; according to experimental testing, the updated Fable 5 completed only 3 of 12 tasks that were previously routine.

The government issued this mandate under the Export Control Reform Act of 2018, which was originally intended to regulate physical hardware exports like uranium enrichment centrifuges. This marked the first instance of export controls being used to limit access to an AI service. Legal scholars note that the order's legal basis is uncertain, as current law typically distinguishes between physical goods and software interactions; previous guidance from the Commerce Department generally considered remote access to U.S.-hosted software to be outside the scope of traditional export controls. Furthermore, the directive was applied unilaterally without the standard public vetting, inter-agency review, or temporary classification processes—such as the 0Y521 code—that typically accompany federal technology restrictions.

Anthropic did not legally challenge the order, characterizing the event as a misunderstanding and opting to negotiate directly with federal officials. This response aligns with public statements from Anthropic CEO Dario Amodei, who argued shortly before the incident that the government should possess the power to block the deployment of dangerous frontier AI models. Additionally, the 2018 statute limits the authority of federal courts to overturn such administrative decisions, requiring challengers to prove a directive is unconstitutional rather than simply unreasonable. This incident signals a shift toward a model of government control where export privileges for AI are conditioned on secretive access to, and modifications of, the most powerful systems currently in development.