The U.S. Department of Commerce officially lifted the export restrictions on the main model of AI startup Anthropic on the evening of June 30, approving full access to Claude Fable5 and Mythos5, which means that this series of models has returned on July 1 after nearly three weeks of being offline.

Previously, due to a potential security vulnerability in Fable5 that could be induced to provide information about network attacks, the U.S. Department of Commerce had issued an emergency export control order on June 12, restricting all foreign citizens, including foreign employees of Anthropic, from accessing the aforementioned models.

QQ20260701-090653.png

With only a 90-minute compliance window, Anthropic was forced to take the model completely offline. After two weeks of intensive negotiations, Anthropic eventually reached a pre-emptive security agreement with the government, promising to proactively detect and resolve security risks associated with the model, collaborate with the government to develop future release agreements, and report malicious activities, thereby avoiding the application process for an export license.

As a simplified version of the Preview mode, Fable5 was previously available to Claude subscribers, while the more sensitive Mythos5 had been approved for targeted distribution to over 100 U.S. organizations last Friday. With the complete revocation of the export ban, Anthropic has confirmed that access will be restored to global users in phases or all at once starting from Wednesday.