US lifts export restrictions on Anthropic's Mythos and Fable models after compliance agreement

The US government has removed export restrictions on Anthropic's Mythos and Fable models, ending a ban that forced the company to cut off public access. Anthropic will begin restoring access to the models on July 1.

On June 12, the US added the models to its export-restricted technologies list, requiring special approval before making them available to foreign nationals. Anthropic found complying with this requirement impractical at scale and suspended public access entirely.

Terms of the agreement

Secretary of Commerce Howard Lutnick stated that Anthropic "has agreed to proactively detect and address security risks associated with the models; to work diligently with the U.S. government on protocols and standards and releases for Mythos, Fable and future models; and to inform the US government of any malicious activity."

According to the company, Anthropic had already publicly pledged to implement many of these measures months before the export rule existed. Cybersecurity experts were skeptical of the restrictions, viewing them as potential political leverage rather than genuine security measures.

Model deployment history

Mythos was initially released to a select group of organizations in April due to concerns about its ability to identify and exploit software vulnerabilities. Fable, a version with additional security guardrails, was released to the public in June before the export ban took effect.

Last week, Lutnick cleared Mythos for release to select customers approved by the White House. OpenAI's latest models were also released to a Trump administration-approved group rather than the public.

Competitive pressure

Asian AI companies have begun releasing models approaching Mythos-level capabilities, including Fugu and Tulonfeng. This development pressured the US government to ease restrictions on Anthropic to maintain American AI competitiveness globally.

Policy uncertainty

The Trump administration's approach to AI policymaking has created uncertainty across the industry regarding future model releases. An executive order issued in June signaling a desire to review models before release drew criticism from analysts including Dean W. Ball, who recently joined OpenAI in a policy position.

What this means

The reversal highlights the tension between AI safety measures and maintaining US technological leadership. The episode demonstrates how export controls—intended as security tools—can become ineffective when companies cannot practically implement them at scale. With competing nations rapidly advancing their AI capabilities, the US faces growing pressure to balance security concerns with the need to keep American AI companies competitive. The lack of clear, consistent policy frameworks creates operational challenges for companies trying to navigate compliance while maintaining market access.