Anthropic suspends Claude Fable 5 worldwide after U.S. export order

Anthropic’s newest frontier model ran headlong into national security limits almost as soon as it reached the public. The company launched Claude Fable 5 and Claude Mythos 5 on June 9, then suspended both models worldwide three days later after a U.S. government export control directive forced it to cut off access for foreign nationals, including its own employees.

Anthropic had introduced Fable 5 as its most capable model and warned that, without safeguards, its cybersecurity abilities could be misused to cause serious damage. To blunt that risk, the company said some cyber-related requests would be routed to Claude Opus 4.8, and that those conservative safeguards would activate in less than 5% of sessions on average. Anthropic also priced Fable 5 at $10 per million input tokens and $50 per million output tokens, underscoring how quickly the model was being pushed into commercial use even as the company flagged the security risks around it.

The shutdown came after Anthropic said it received the directive at 5:21 p.m. ET on June 12. The order applied to any foreign national, whether inside or outside the United States, including foreign-national Anthropic employees. Because the company said it could not reliably separate affected users in real time, it disabled Fable 5 and Mythos 5 for all customers worldwide to ensure compliance. All other Anthropic models remained available.

The models were not ordinary public releases. Anthropic said Fable 5 and Mythos 5 were intended for a trusted-access program and were being deployed in Project Glasswing, a U.S. government collaboration focused on securing critical software. That detail makes the suspension more consequential than a product rollback. It shows how a model can move from launch to restricted access in days when government concerns over capability, nationality, and export rules collide.

Anthropic has been sharpening its safety posture for months. Its June 2 policy update warned of malware creation, scaled abuse, and cyberattacks as risks from agentic AI, and the company has separately added safeguards around cybersecurity and biology. It has also publicly argued against uses it views as dangerous, including mass domestic surveillance and fully autonomous weapons. The episode now stands as a test case for the broader race to release more capable models, where innovation can be slowed not only by technical limits but by the reach of national security policy.