What happened
On June 12, 2026 at 5:21 PM ET, Commerce Secretary Howard Lutnick issued an export control directive to Anthropic CEO Dario Amodei invoking national security authorities under export control law. The directive suspended all access to Claude Fable 5 and Mythos 5 models for any foreign national, whether inside or outside the United States, including foreign Anthropic employees. Anthropic disabled both models globally for all customers to ensure compliance. The action cited concerns about a potential jailbreak or bypass of Fable 5's safety guardrails reported by a third party.
Why it matters
This represents an unprecedented use of export controls against commercially deployed frontier AI models, marking a major escalation in national-security-driven AI model restrictions. The directive forcibly took advanced AI capabilities offline globally within 90 minutes of government notification, setting a precedent that governments can unilaterally restrict AI model access based on national security assertions. This affects international AI competition, cross-border AI development collaboration, and establishes that model access itself—not just training infrastructure—is now treated as a controlled export subject to licensing requirements.
Action needed
AI labs must monitor Commerce Department communications for export control directives and establish rapid compliance protocols. Organizations relying on frontier models should assume model availability can be suspended with minimal notice for national security reasons. Procurement and service-level agreements should account for potential forced model discontinuation. Monitor Commerce Department and BIS guidance for clarification of export control criteria for AI models.