Trump Administration Lifts Export Ban on Anthropic's Most Powerful Models After Three-Week Standoff

US Commerce Secretary Howard Lutnick announced the lifting of export controls on Anthropic's two most powerful models, Claude Fable 5 and Mythos 5. The decision ends a three-week dispute that cut off foreign customers from the company's most advanced tools and raised questions about who in Washington actually calls the shots on the fate of American artificial intelligence.

The whole affair began on June 12, just three days after the public launch of Fable 5 and Mythos 5. Andy Jassy, Amazon's CEO, contacted Treasury Secretary Scott Bessent and AI adviser David Sacks directly, reporting that Amazon researchers had found a way around Mythos 5's safeguards. A specific prompt sequence reportedly allowed users to extract cyberattack-related information from the model that should have been restricted.

Ninety Minutes to Decide

The Commerce Department reacted instantly. Dario Amodei, Anthropic's CEO, was given just 90 minutes to take the models offline for users outside the US. The unprecedented speed showed how anxious the Trump administration is about national security issues tied to cutting-edge AI systems, especially when the alarm comes from a rival in the cloud business.

On June 26 the situation was partially eased with limited access restored for more than a hundred US companies under a program codenamed Project Glasswing. That did not solve the core problem: for nearly three weeks, customers outside the United States had no access to Anthropic's most capable models, while competitors, including Chinese firms offering cheaper and less restricted alternatives, were free to grab market share.

Experts Push Back

Not all cybersecurity experts shared the administration's alarm over the scale of the threat. Andrew Morris of GreyNoise Intelligence said the disclosed information was still far from genuinely dangerous. Dean Ball, an AI policy expert, called the Commerce Department's response a caricature. Researcher Ben Murphy, meanwhile, noted that the whole incident marks another step toward the balkanization of AI technology along geopolitical lines.

Over the past two weeks we worked closely with Anthropic to review and clear Fable 5, ensuring alignment across the US government and strengthening American leadership in AI - Howard Lutnick, US Secretary of Commerce

Terms of the Truce

Resolving the dispute does not mean a return to the pre-June 12 status quo. Anthropic agreed to proactively detect and remove security risks tied to its models, blocking the flaw Amazon found in 99 percent of cases. The company also committed to closer cooperation with the federal government on security protocols and future model launches, and to reporting any malicious activity it detects among users to authorities. Anthropic is additionally set to build a joint vulnerability-assessment system together with Amazon, Microsoft and Google.

Worth noting is the change at the head of the negotiations on Anthropic's side. It was not Dario Amodei, known for his public calls for AI regulation and his earlier support for Kamala Harris in the 2024 election, who led talks with the administration, but company co-founder Tom Brown. Lutnick's letter was addressed to Brown directly, which commentators read as a signal that Amodei had become a political target for the administration.

What It Means for Firms Outside the US

For companies and developers outside the United States, including Polish ones, this episode is a warning that access to the most advanced AI models can be cut off overnight by administrative decision, regardless of contracts or deployment plans. The three-week disruption in access to Fable 5 and Mythos 5 pushed some customers to look for alternatives, including from Chinese vendors, underscoring how fragile supply chains can be when they depend on a single provider subject to US export policy.

Anthropic announced the restoration of global access to both models starting July 1, telling customers the unlocking process would be gradual. The company also published an official statement on its news site describing both the June 12 suspension and the subsequent restoration of access.

The matter also reached higher diplomatic levels. According to reports, a meeting between Dario Amodei and Trump administration officials on the sidelines of the G7 summit in France helped break the negotiating deadlock, where the future of American AI leadership amid growing competition from China was also on the agenda.

The whole incident also highlights growing tension between leading AI companies and the US government over who bears responsibility for the security of the newest models. While Anthropic has spent months positioning itself as the safety-focused AI company, the fact that its own model triggered an alarm over a possible leak of cyberattack information may complicate that narrative in the eyes of institutional and government customers worldwide.

Sources: Trump administration lifts export controls on Anthropic's most powerful AI models (Yahoo News), Inside Anthropic's 20-Day Standoff With the Trump Administration Over Pulled AI Models (MLQ News), Trump administration lifts Claude Mythos 5, Fable 5 export restrictions after Anthropic works with government (Fox Business), Trump Just Cleared Anthropic to Go Global (24/7 Wall St.)