US export controls on AI reversed in under 3 weeks — was it overreach?

The US export controls on AI systems came and went in less than three weeks — but the implications of that short, turbulent window are still reverberating across the technology industry. On July 1, Anthropic confirmed that the US Department of Commerce had fully lifted restrictions on its two most advanced models, Claude Fable 5 and Mythos 5, ending a saga that exposed serious cracks in how Washington regulates cutting-edge AI software.

Key takeaways

• The US Department of Commerce lifted export controls on Anthropic’s Fable 5 and Mythos 5 on July 1, less than three weeks after imposing them.
• Commerce Secretary Howard Lutnick imposed the restrictions around June 12–13, citing national security concerns over potential misuse by adversarial actors.
• The initial ban blocked access globally, including for Anthropic’s own employees, before a partial lift on June 27 restored access to over 100 vetted companies.
• Anthropic agreed to proactively detect security risks, collaborate with the government on future model releases, and report any malicious activity.
• No crypto tokens or DeFi protocols were affected by these controls.

How a national security order shut down a commercial AI product worldwide

When Commerce Secretary Howard Lutnick signed the export-control order around June 12–13, the fallout was almost immediate — and far broader than most observers expected. Anthropic had no clean way to restrict access only to foreign nationals. The result was a total global suspension of both Fable 5 and Mythos 5, cutting off not just overseas users but the company’s own staff.

The move was framed around national security. US authorities had grown concerned that the models’ capabilities — particularly Mythos 5, which was built for businesses and cybersecurity experts and is designed to identify and exploit vulnerabilities in computer code — could be misused by military or intelligence actors in China, Russia, or other countries of concern. Fable 5, aimed at consumers and capable of deep reasoning and complex autonomous tasks, was swept up in the same order. Both had only been released on June 9, just days before the suspension.

Anthropic pushed back at the time, arguing that US authorities had not identified specific concerns about its technology. The company said its understanding was that the government believed a method of bypassing Fable 5’s safety restrictions — a so-called jailbreak — had been found, but Anthropic disputed that a narrow potential jailbreak in a commercially deployed product warranted a full recall.

A partial reversal, then a full one

The total blackout held for roughly two weeks. On June 27, the US government offered a partial reprieve, granting access to over 100 vetted companies and institutions, many of them Fortune 500 firms, without requiring individual export licenses. That partial restoration was telling: it suggested the initial order had been applied more broadly than the underlying security concern actually justified.

By July 1, full access was restored. In a letter to Anthropic seen by both the BBC and Reuters, Lutnick confirmed that the export controls were withdrawn and that a license was no longer required for the models’ export. The Commerce Department reserved the right to revisit the decision if circumstances changed.

In exchange, Anthropic agreed to proactively detect and address security risks associated with both models, work with the US government on protocols and standards for Mythos, Fable, and future releases, and alert authorities to any malicious activity detected. “We’ll begin restoring access tomorrow,” the company said in a statement on X following the announcement.

The regulatory overreach at the center of this story

The deeper issue this episode surfaced is a structural one. The Bureau of Industry and Security, the Commerce Department division responsible for export controls, was applying a regulatory framework designed for physical goods — semiconductors, missile components, industrial machinery — to a cloud-based AI service. The toolkit simply wasn’t built for this.

The result was an overcorrection. Allied nations lost access alongside adversarial ones. Domestic commercial users were locked out. Even the company that built the models couldn’t use them internally. When the framework for restricting a tangible export item gets applied to a software API, the blast radius becomes impossible to contain with precision.

The episode wasn’t unique to Anthropic. OpenAI delayed the full public launch of GPT-5.6 at the US government’s request, limiting access to a small group of vetted partners around the same period. OpenAI CEO Sam Altman commented publicly that extensive safety testing was not inherently objectionable, but that he was uncomfortable with the government choosing which customers get access. That tension — between legitimate national security oversight and open commercial deployment — is now squarely on the table for the entire AI industry.

What the tiered restoration reveals about where AI regulation is heading

The phased rollback is worth examining closely. The fact that authorities could rapidly identify and approve over 100 vetted companies for partial access on June 27, only to grant full public access days later, suggests the government is already thinking about AI access in tiers. Organizations with existing government relationships, compliance infrastructure, and the resources to navigate a vetting process were back online while smaller developers and individual users waited.

That structure has real competitive implications. If future export control frameworks for AI follow a similar pattern, large enterprises with the capacity to engage the Bureau of Industry and Security will have continuity of access. Startups, independent researchers, and developers in smaller markets may face gaps that compound over time.

For its part, Anthropic has now established a formal security collaboration channel with the US government — one that will carry forward to future model releases. That relationship could be an asset in navigating future restrictions, or it could set expectations that become difficult to manage as models grow more capable.

No impact on crypto or digital assets

For those tracking the intersection of AI and digital assets, the Anthropic export control episode carried no direct consequences. No crypto tokens were affected, and no DeFi protocols were disrupted by the restrictions or their removal. Anthropic has no associated token, and the controls touched only the company’s commercial AI models.

FAQ

Why were export controls imposed on Anthropic’s AI models?

Export controls were imposed around June 12–13 by Commerce Secretary Howard Lutnick, citing national security concerns. US authorities were worried the models could be misused by military or intelligence actors in adversarial countries, with particular concern around a reported jailbreak method for Fable 5.

Who was affected by the initial export restrictions?

The restrictions blocked access globally for all users, including foreign nationals and Anthropic’s own employees. Because Anthropic had no mechanism to restrict only foreign access to its cloud-based service, it was forced to suspend the models entirely worldwide.

How long did the export control restrictions last?

The total restriction and restoration cycle lasted under three weeks, running from the order around June 12–13 to partial restoration on June 27 and full reinstatement on July 1.

Did the export controls impact any crypto tokens or DeFi protocols?

No. The export controls applied specifically to Anthropic’s AI models Fable 5 and Mythos 5. No crypto tokens or DeFi protocols were affected by these restrictions or their removal.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.