The Trump administration, through the Commerce Department, has directed Anthropic to obtain U.S. government approval before allowing foreign nationals to use its Fable 5 and Mythos 5 AI models.
The order represents an expansion of export control authority to cover access and usage of frontier AI models, rather than only the transfer of technology or hardware.
Key Details of the Order
Affected models: Fable 5 and Mythos 5.
Requirement: Anthropic must secure U.S. approval for foreign nationals to access or use these models.
Legal mechanism: The Commerce Department is applying export control laws in a novel way, targeting theusageof advanced AI systems by non-U.S. persons.
Scope: This goes beyond traditional controls on exporting chips or model weights and extends to controlling who can interact with the models themselves.
Rationale (per reporting): Part of a broader effort to rein in access to cutting-edge AI capabilities amid national security concerns.
The move has drawn attention for testing the outer boundaries of existing export control statutes and raising questions about the government’s ability to regulate access to AI systems rather than just their physical or technical export.
Actions to Take
AI labs and frontier model companies: Review foreign user access policies and data flows for high-capability models. Prepare compliance processes for potential similar orders on other frontier systems.
Enterprise customers using Anthropic models: Audit which teams or regions have access to Fable 5 or Mythos 5. Map potential disruptions if additional approval requirements or restrictions are imposed.
Legal and compliance teams at AI companies: Assess exposure under expanded interpretations of export control rules. This order may set precedent for other model providers.
Investors and boards of AI companies: Factor increased regulatory and compliance risk into valuations and go-to-market plans for non-U.S. customers, particularly in sensitive verticals or geographies.
Policy and government affairs teams: Monitor for follow-on actions targeting other frontier labs or models. The use of export controls on modelusagerather than transfer is a significant expansion.
This is one of the more direct regulatory interventions to date aimed at controlling access to specific frontier AI models by foreign nationals.