US Lifts Export Curbs on Anthropic’s Mythos and Fable AI Models

The U.S. government has removed the requirement for Anthropic to obtain an export license before making its Mythos and Fable AI models available internationally. The restriction had effectively blocked public access to what many consider the most advanced AI models currently released.

Anthropic confirmed that it will begin restoring public access to both models on Wednesday, July 1.

On June 12, the U.S. government placed Mythos and Fable on its export-controlled technologies list, preventing the company from providing access to foreign nationals without obtaining special authorisation. Because enforcing the licensing requirement at a global scale proved impractical, Anthropic suspended public access to the models altogether.

Following several weeks of discussions, U.S. Secretary of Commerce Howard Lutnick announced that Anthropic had agreed to strengthen its cooperation with federal authorities. According to Lutnick, the company “has agreed to detect and address security risks associated with the models proactively; to work diligently with the U.S. government on protocols and standards and releases for Mythos, Fable and future models; and to inform the U.S. government of any malicious activity.”

Much of this cooperation had already been publicly promised by Anthropic months before the export controls were introduced. That was one reason many cybersecurity experts questioned the restrictions from the beginning. Critics argued the move appeared less focused on national security and more like political leverage against Anthropic after several company executives publicly criticised how the government—and political opponents of President Trump—might ultimately use advanced AI technology.

Anthropic initially provided Mythos to a limited group of organisations, beginning in April, to address concerns about the model’s ability to identify and exploit software vulnerabilities. A separate version, known as Fable, became publicly available in June with additional security safeguards designed to reduce potential misuse.

At the same time, AI developers across Asia have continued narrowing the technological gap. Companies behind models such as Fugu and Tulongfeng have introduced systems that reportedly approach Mythos’s capabilities. The increasing international competition placed additional pressure on the U.S. government to relax export controls and allow American AI companies to remain competitive in global markets.

Last week, Lutnick authorised Mythos for deployment to a select group of organisations approved by the White House. Around the same time, OpenAI’s latest models were also made available only to organisations authorised by the Trump administration, rather than being released to the broader public.

The administration’s shifting approach to AI regulation has created continued uncertainty across the technology industry regarding future model approvals and release requirements. An executive order issued in June that suggested additional government review before advanced AI models could be released drew criticism from several policy experts, including Dean W. Ball, who recently accepted a policy role at OpenAI.