What happened
On June 12, 2026 (received by Anthropic at 5:21 PM ET), Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei placing Claude Fable 5 and Mythos 5 under export controls, prohibiting access by any foreign national — including foreign-born Anthropic employees — whether inside or outside the United States. Because full compliance was not achievable without disabling the models entirely, Anthropic shut down both models for all customers. The government's stated rationale was an alleged jailbreak technique that could allow Fable 5 to identify software vulnerabilities, though Anthropic disputed this characterisation, stating it found the capability was already available in other public models including OpenAI GPT-5.5.
Why it matters
This is the first time the US government has applied export controls directly to an AI model's access API rather than to chips or hardware, establishing a precedent that frontier model access is a controllable national security asset. It exposes a previously unquantified enterprise risk: any AI service — regardless of pre-existing contracts or integrations — can be abruptly disabled by government directive on jailbreak-related grounds with no advance notice, disproportionately affecting multinational organisations and any enterprise with foreign-national employees or users.
Action needed
AI-dependent organisations should immediately audit their AI vendor contracts for force-majeure clauses covering government directives, assess their substitutability posture (which model-agnostic fallback paths exist if a primary model is pulled), and review internal AI access policies for employees who are foreign nationals to ensure compliance precedents are understood.