U.S. Government Halts World's Most Advanced AI Deployment
The Trump administration forced Anthropic to disable its most powerful AI models worldwide, marking the first time federal export controls took a frontier system offline and demonstrating American authority over global technology infrastructure.
The Trump administration pulled the plug on the world's most advanced artificial intelligence Friday, forcing Anthropic to disable its flagship models globally through export controls that left foreign governments, companies and researchers locked out of systems they had only days before been using.
Commerce Secretary Howard Lutnick ordered the immediate suspension of Fable 5 and Mythos 5 for all foreign nationals at 5:21 p.m. ET June 12. The directive compelled a worldwide shutdown that rippled across continents, cutting off access from London to Tokyo.
Anthropic could not separate foreign users from domestic customers in real time, so the San Francisco company disabled both models for all customers worldwide. The order followed a report of a jailbreak involving codebase flaw detection, though the government provided only verbal evidence of the vulnerability.
Anthropic later described the alleged vulnerability as a "narrow, non-universal jailbreak." The action marks the first known instance of the U.S. government forcing a publicly deployed frontier AI model offline, proving federal authority supersedes corporate interests regardless of commercial impact.
The shutdown represents the latest escalation in a months-long dispute between Anthropic and the Trump administration. The Pentagon designated Anthropic a "supply chain risk" in early March after the company refused to allow unrestricted military use of its models for autonomous weapons and mass surveillance.
Defense Secretary Pete Hegseth had given CEO Dario Amodei a Friday deadline in February to comply with Pentagon demands. Anthropic rejected the order, citing ethical restrictions.
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic stated June 12. The company argued the same capability exists in other public models, including OpenAI's GPT-5.5, and is "used every day by the defenders who keep systems safe."
Anthropic said it is complying with the government directive but disagreed with the finding. The company stated the action did not adhere to principles of a transparent, fair, clear and technically grounded statutory process, then apologized for the disruption to customers.
Pentagon CIO Kirsten Davies immediately countered the company's complaints. "Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation," Davies stated. "America First. Always."
The Defense Department's stance reflects growing concerns about Silicon Valley prioritizing product releases over national security, particularly as China actively steals American technology.
Amazon Web Services revoked access to Fable 5 and Mythos 5 globally. The models had launched just four days earlier on June 9, priced at $10 per million input tokens and $50 per million output tokens.
Stripe had already put Fable 5 to work, compressing months of engineering work on a 50-million-line Ruby codebase into days. The worldwide shutdown demonstrates the U.S. government's absolute authority over frontier AI deployed by American companies, regardless of foreign reliance or commercial disruption.
Silicon Valley's complaints about the export controls ignore the geopolitical reality of American technological supremacy and the threat of foreign espionage.
"If you describe your product as a munition in every press release, eventually a government takes you at your word," said Peter Girnus, senior threat researcher at Zero Day Initiative. AI critic Gary Marcus warned the government action would likely convince many Chinese-born AI researchers currently working for American labs to return to China.
Anthropic's immediate compliance with the June 12 directive contradicts its prior public stance of resisting government overreach. The company had sued the federal government over the supply chain risk designation, winning a preliminary injunction from Judge Rita Lin on March 26.
The D.C. Circuit denied a stay April 8, allowing the designation to remain while legal battles continue.
The shutdown comes as Anthropic prepares for an initial public offering after a May funding round at a $965 billion valuation, followed by a confidential SEC filing on June 1. The company's revenue rate topped $19 billion annually as of March, with Fable 5 and Mythos 5 representing its most advanced and expensive offerings.
Models remain offline as the Commerce Department reviews the national security implications of the alleged jailbreak.
Dean W. Ball, a senior fellow at the Foundation for American Innovation and former Trump administration official, called the action "simply cartoonish." He questioned why the administration would block Britain and other allies from using American AI models while continuing to export advanced AI chips to China.
The irony exposes Silicon Valley's selective outrage over national security measures while ignoring the real threat of Chinese technology theft.
The "narrow jailbreak" rationale, combined with the Pentagon's prior designation, validates the government's national security concerns over Anthropic's cybersecurity-focused models. Amodei himself had called for FAA-style mandatory safety testing in a June 10 essay, arguing government should have power to block unsafe frontier AI deployments.
The Commerce Department action demonstrates that authority already exists and will be exercised when national security demands it.
Anthropic continues to work with U.S. government partners through its Project Glasswing program, which included Amazon, Apple, Google, Microsoft and CrowdStrike among approximately 50 vetted organizations. The company red-teamed its models for thousands of hours with government agencies including the UK's Artificial Intelligence Security Institute.
An external bug bounty produced no universal jailbreaks in over 1,000 hours of testing. The UK AISI made progress toward a universal jailbreak within a brief initial testing window.
The models' sudden disappearance from global markets underscores America's technological dominance in the AI race. While Silicon Valley complains about export controls, the U.S. government has demonstrated it can dictate terms worldwide through control of American companies' infrastructure.
AWS's global revocation of access proves that when the United States acts, the world's most advanced AI systems become inaccessible beyond its borders. The people who built them cannot keep them running once Washington speaks.