After a dramatic two-week global shutdown, Anthropic’s most advanced AI models are back online. Commerce Secretary Howard Lutnick announced on June 29 that the U.S. government has lifted its export control order on Fable 5 and Mythos 5, ending a standoff that paralyzed critical services and sparked frantic negotiations.
The reversal allows Anthropic to reactivate its models for over 100 U.S. companies and government agencies. The decision followed an intense, high-stakes debate over national security that pitted a leading AI lab against federal regulators.
Anatomy of the Ban
The drama began on June 12, when the Department of Commerce issued a directive forcing Anthropic to cut off access to its most powerful models for all foreign nationals, citing national security authorities.
Unable to reliably filter users by citizenship in real-time, Anthropic was forced into a full global shutdown. This move instantly hobbled enterprise clients in finance, healthcare, and critical infrastructure who depend on the models through platforms like AWS Bedrock, Google Cloud, and Microsoft Foundry.
The government’s aggressive order was reportedly triggered by a rival company's claim to have "jailbroken" Mythos 5. This raised alarms about its potential use in offensive cyber operations, prompting the swift federal intervention.
An Uneasy Truce
Secretary Lutnick confirmed the ban was lifted after Anthropic agreed to collaborate with the government on new safety protocols and a review process for future model releases. This agreement forges a potential new playbook for how Washington will oversee powerful, dual-use AI systems.
Throughout the shutdown, Anthropic argued the government’s action was a "misunderstanding" of a narrow technical issue. The company found allies in the infosec community, with experts publishing an open letter demanding the directive be reversed.
The government dropped the ban just three days after Fable 5 and Mythos 5 had been publicly launched. Mythos 5, in particular, was marketed as a next-generation cybersecurity tool for defenders, but its dual-use potential for creating major cyber intrusions was clear from the start.