Anthropic: Trump Administration Lifts Export Controls on Claude Fable 5 and Mythos 5

Anthropic announced Tuesday that the U.S. Department of Commerce has rescinded export controls on its Claude Fable 5 and Mythos 5 artificial intelligence models. This development marks the resolution of a recent, high-profile dispute between the AI firm and the Trump administration.

“We are immensely grateful for the patience of our users and for the collaborative efforts of all those who worked with us to redeploy these models,” Anthropic stated in a social media update.

In mid-June, Anthropic proactively suspended access to Fable 5 and Mythos 5 to comply with a government directive concerning export controls. The directive cited “national security authorities” and mandated the suspension of all access by “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.”

This governmental action against Anthropic occurred against a backdrop of a rapid ascent in the capabilities of Chinese open-source AI models. These models have demonstrated performance comparable to leading U.S. offerings, often at a significantly lower cost. Concerns were raised by numerous tech executives and investors that the administration’s restrictions on Anthropic’s model deployment could inadvertently grant Chinese developers a crucial advantage in bridging the AI development gap.

Tuesday’s announcement follows closely on the heels of U.S. Commerce Secretary Howard Lutnick granting Anthropic permission to release Mythos 5 to a select cohort of companies and federal agencies. In a letter to Anthropic, reviewed by this publication, Lutnick indicated he had determined that “appropriate safeguards” were in place, enabling specific “trusted partners” to access the model.

“Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. Government and strengthen America’s leadership in AI,” Lutnick communicated in a social media post on Tuesday.

Anthropic confirmed that access to Fable 5 would begin to be restored on Wednesday.

The company had launched these advanced models, positioning them as industry-leading across various benchmarks, just days prior to the export control directive. Fable 5, in particular, represented Anthropic’s first public release of such a sophisticated offering.

Anthropic, which has engaged in considerable dialogue with the White House this year, swiftly initiated negotiations with the Trump administration in Washington, D.C., following the receipt of the export control directive on June 12. The company maintained a discreet stance in the subsequent weeks, offering limited public updates on the progress of these discussions.

Reports indicate that Tom Brown, an Anthropic co-founder, took the lead in negotiations with the Trump administration, assuming this responsibility from CEO Dario Amodei. Amodei had reportedly become a focal point for the administration due to his outspoken advocacy for AI safety and his prior public support for Kamala Harris during the 2024 presidential election.

Notably, Lutnick’s correspondence on Friday was directed to Brown, rather than Amodei.

This episode has generated considerable uncertainty within the AI industry. Industry leaders have expressed a lack of clarity regarding the government’s regulatory trajectory and the key influencers shaping presidential decisions, especially following the departure of David Sacks from his role as crypto and AI czar earlier in the year.

The administration has signaled an intent to play a more assertive role in AI regulation. President Donald Trump signed an executive order on AI in early June. This order, which was broad in its directives, called for AI developers to voluntarily submit their models for government assessment of capabilities prior to their full release, and it mandated federal agencies to establish relevant frameworks and processes within 60 days.

This directive has presented challenges for AI companies, including OpenAI, in managing their model launch schedules.

On Friday, OpenAI unveiled three new models, stating its compliance with the government’s request to initially limit the rollout to a “small group of trusted partners.” OpenAI indicated that it had previewed the capabilities of these models, including GPT-5.6, and shared its deployment plans with the government in advance of the Friday launch.

“We do not believe that this type of government access process should become the long-term default,” OpenAI communicated in its Friday statement. “It restricts access to essential tools for users, developers, enterprises, cybersecurity professionals, and global partners who require them.”