U.S. Lifts AI Export Restrictions on Anthropic Models After Global Shutdown

The June 12 Shutdown and Regulatory Intervention

On June 12, 2026, the Department of Commerce issued a directive requiring Anthropic to prevent all foreign nationals from accessing Claude Fable 5 and Mythos 5. Because the company could not immediately verify the nationality of its global user base, it disabled both models globally within hours.

The order relied on the Export Control Reform Act of 2018, a law originally designed for hardware like uranium enrichment centrifuges. Legal scholars noted that this was the first time the government used such export controls to restrict access to an AI service, raising questions about whether a digital response to a prompt constitutes an “export” under existing federal law.

Safety Guardrails and the Return of Access

Following the cutoff, Anthropic worked closely with the White House to address vulnerabilities in the models’ safety systems. On June 30, Secretary of Commerce Howard Lutnick formally withdrew the emergency restrictions, stating that the administration was satisfied with the steps taken to mitigate risks.

“We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We’ll begin restoring access tomorrow.”

Anthropic, via post on X

While access to Fable 5 was restored globally on July 1, VentureBeat reported that Mythos 5 remains restricted to a vetted-access model. Anthropic continues to manage Mythos 5 through its Project Glasswing cybersecurity testing program, keeping the model behind a gate even though the broader emergency export order has been lifted.

For more on this story, see U.S. Grants Iran Massive Oil Sanctions Waivers: How Billions in Revenue Could Reshape Tehran’s Nuclear & Global Strategy” (Alternative options if needed:) “U.S. Waives Iran Oil Sanctions-But Will Tehran Comply? Trump’s Warning & the Nuclear Standoff Explained” “Breaking: U.S. Lifts Iran Oil Sanctions-Billions Unlocked, But Nuclear Talks Remain in Limbo” “Iran Sanctions Waiver: How the U.S. Move Could Fuel Tehran’s Economy & Nuclear Ambitions.

The Strategic Debate Over AI Dominance

Critics of the government’s “ad hoc” approach, including former Facebook security chief Alex Stamos, warned that forcing companies to disable their most advanced tools could drive security firms toward alternatives developed in China.

Anthropic has framed the competition differently, emphasizing that the U.S. must defend its compute advantage to prevent authoritarian regimes from shaping global AI norms. In a research paper released in 2028, the company noted that while Chinese AI labs possess world-class talent, they remain constrained by limited access to the most capable American-made chips. The company expressed concern that if U.S. firms are hobbled by overly restrictive regulations, it could unintentionally accelerate the ability of rival regimes to catch up.

Financial and Operational Costs

For enterprise customers, the return of Claude Fable 5 comes with a significant price tag. Anthropic currently prices both Fable 5 and Mythos 5 at $10.00 per million input tokens and $50.00 per million output tokens.

Model	Input ($/1M)	Output ($/1M)
Fable 5 / Mythos 5	$10.00	$50.00

Susie Wiles, the president’s chief of staff, indicated that the administration’s focus remains on testing and safety, though the specific mechanisms for future model releases are still evolving. As the industry moves forward, the primary uncertainty lies in how the government will balance its desire for rapid, safe deployment with the legal and operational realities of controlling software that resides on domestic servers but serves a global audience.