US Government Imposes Model Access Restrictions on Anthropic Citing National Security
The U.S. government ordered Anthropic to immediately suspend access to its advanced AI models Claude Fable 5 and Mythos 5 for all foreign nationals, citing national security concerns. This represents a significant precedent in AI model export control and raises questions about the enforceability and scope of such restrictions.
Affected
Anthropic received a government order at 17:21 ET instructing immediate suspension of access to its most capable models for foreign nationals regardless of geographic location. The company chose to disable these models entirely rather than implement geographic or identity-based restrictions, suggesting either technical infeasibility of granular controls or a deliberate business decision to avoid the complexity of compliance.
This action represents a significant escalation in AI governance. Unlike traditional export controls that restrict physical goods or technical documentation, this targets runtime access to a service. The order's breadth matters: restricting foreign nationals globally prevents legitimate international collaboration and research. The timing of a full suspension rather than a phased rollout indicates urgency on the government's part regarding what specific capabilities or potential dual-use applications may be at stake.
The national security rationale is not detailed in the available reporting, but plausible concerns include: advanced reasoning capabilities that could accelerate weapons development, code generation at scale for critical infrastructure systems, or novel social engineering or information operations capabilities. The fact that both Fable 5 and Mythos 5 were targeted suggests either identical capability profiles or a broader sweep of all cutting-edge models.
Organisations deploying Anthropic models for international teams face immediate operational disruption. The enforceability mechanism is unclear: Anthropic could theoretically re-enable access in defiance, but faces potential liability or enforcement action. This creates a chilling effect for other AI vendors considering resistance to similar orders.
The strategic implication is substantial. U.S. AI leadership depends on attracting international talent and maintaining access to global compute resources and datasets. Unilateral restrictions without clear technical justification risk accelerating rival AI development programmes and fragmentation of the global AI ecosystem. The precedent may also invite reciprocal restrictions from other nations.
Sources