The U.S. Department of Commerce has issued new export restrictions on Anthropic’s AI models, Claude Mythos 5 and Claude Fable 5, marking a significant move to secure national interests. The Bureau of Industry and Security (BIS) communicated this directive to Anthropic CEO Dario Amodei, requiring the company to secure an individual export license before sharing these models with foreign nationals globally.
Regulatory Framework and Security Concerns
Commerce Secretary Howard W. Lutnick, in his letter, referenced the Export Control Reform Act of 2018 (ECRA), specifically section 50 U.S.C. § 4817(b)(1). This empowers the BIS to apply temporary controls on technologies vital to U.S. national security. Additionally, the Export Administration Regulations (EAR) under § 744.22(b) support this move, mandating licenses when there is potential military-intelligence application.
This decision marks the first instance of the Commerce Department leveraging this ECRA clause against a commercially available AI model, amidst concerns over potential misuse of these technologies.
Comprehensive Coverage of Restrictions
The BIS order impacts any export, reexport, or internal transfer of the Claude Mythos 5 and Fable 5 models. This includes taking or sending the models out of the U.S., transferring them between foreign nations, or even releasing them to foreign individuals within the U.S., including non-citizen employees of Anthropic.
Anthropic was notified of this requirement on June 12, 2026, just days post-launch of these models. The company promptly disabled international access to both AI systems in compliance with the directive.
Industry Reactions and Future Implications
The action stems from a reported vulnerability in Fable 5, which allegedly allows breaches of its safety protocols, posing significant cybersecurity risks. Mythos 5, capable of identifying software flaws, was initially designed for government and select enterprise use. Anthropic argues the vulnerability is minor and can be found using other available models.
More than 80 tech leaders, including Nvidia and Adobe executives, have urged the U.S. government to reconsider, emphasizing that the perceived threat may be exaggerated. In response to these controls, Anthropic is required to file for licenses through the SNAP-R portal, adhering to stringent compliance measures.
These export controls could strain U.S. diplomatic relations, with the EU Commission cautioning against discriminatory practices. The restrictions also pose challenges for foreign nationals on H-1B visas in the AI sector, potentially impacting workforce dynamics.
The BIS indicated that these license mandates will remain until officially revised or revoked. This precedent highlights the U.S. government’s approach to regulating advanced AI technologies under dual-use export control laws.
