US export controls suspend Anthropic's Claude Fable 5, Mythos 5 after 3-day launch window

Anthropic launched Claude Fable 5, its new Mythos-class foundation model, on June 9, 2026. Three days later, on June 12, the company received a US government export-control directive requiring it to suspend access to both Fable 5 and Mythos 5 for all customers immediately. The ban applied only to the new Mythos-class tier; all other Claude models (Opus, Sonnet, Haiku) remain accessible.

Fable 5 was positioned as a generally available, safety-guardrail version of the Mythos-class tier, designed for coding, long-horizon agentic work, vision tasks, and complex research workflows. Mythos 5 (restricted to limited availability under Project Glasswing) lacks some of those safeguards and offers enhanced capabilities in areas flagged by U.S. regulators—particularly cybersecurity and offensive capability. The 72-hour window between public launch and suspension reflects newly tightened administration policy on frontier AI model export controls.

Anthropic stated it is complying with the order, disabling both models for all customers globally. The company did not disclose the specific legal or technical justification cited by regulators, though prior months saw mounting scrutiny over model capabilities in offensive cybersecurity and red-teaming scenarios. This marks one of the earliest government-level model bans targeting a commercial frontier lab.

Architects and enterprises that began integrating Fable 5 will need to revert to Claude Opus 4.8 or design around the Mythos-class tier withdrawal. The incident signals U.S. regulators are willing to act unilaterally on frontier models post-launch if capabilities or export concerns emerge. Future model releases will likely face pre-launch security review; practitioners should account for policy risk in AI procurement planning.