US export controls force Anthropic to suspend Claude Fable 5 for all users

On June 12, 2026, just three days after its general availability launch on June 9, Anthropic suspended access to Claude Fable 5 and Claude Mythos 5 in response to a U.S. government export-control directive. The suspension applies globally to all users, not just foreign nationals, because Anthropic cannot gate access by nationality in real time without rebuilding auth infrastructure. Fable 5, Anthropic's first Mythos-class model made safe for public use, had achieved 95.0% on SWE-bench Verified and topped coding leaderboards on the day of launch before the directive forced it offline.

Anthropic stated access is expected to return around July 1, 2026, but only for U.S.-based users initially. Claude Mythos 5, the unrestricted variant without safety classifiers and available only to Project Glasswing partners, remains suspended with no restoration timeline. All other Claude models (Opus 4.8, Sonnet 4.6, Haiku 4.5) are unaffected. The suspension marks the first major enforcement of post-Biden AI export controls on a frontier model, following months of interagency debate over which AI capabilities constitute controlled items. Fable 5 cost $10/$50 per million tokens (double Opus 4.8) and was positioned as Anthropic's cutting-edge long-horizon reasoning and scientific research capability.

For AI product teams, this signals that export controls on frontier models are now operational and can be enforced with minimal advance notice. The ~3-week blackout window highlights the compliance cost of rapid international model launches: architecture for geofenced access and rapid deployment must be built into CI/CD pipelines if companies intend to navigate different regulatory regimes. Stripe's early testing reported Fable 5 could compress a two-month codebase migration into a single day—a capability loss that affects engineering orgs relying on cutting-edge reasoning in production systems.