U.S. Orders Anthropic to Disable Advanced AI Models Over Security Concerns

In an unprecedented escalation of technological containment, the U.S. government has ordered artificial intelligence company Anthropic to immediately suspend access to its most advanced AI models for all foreign nationals.[1][7]

The emergency export control directive, issued by the Commerce Department late Friday afternoon, targeted Anthropic's newly released Fable 5 and Mythos 5 systems. Because the order applies to any non-U.S. citizen—including Anthropic's own foreign-born employees and users residing within the United States—the company concluded that selective compliance was impossible.[2][4][6]

As a result, Anthropic abruptly disabled both models globally for all customers, pulling its flagship products offline just three days after their highly anticipated launch.[1][2]

The Commerce Department cited unspecified national security concerns in its letter to Anthropic CEO Dario Amodei. According to the company, the government acted on a tip from a third party claiming to have found a method to "jailbreak" Fable 5, bypassing its safety guardrails to identify software vulnerabilities.[1][5][6]

Anthropic fiercely contested the government's rationale, calling the situation a "misunderstanding." In a public statement, the company argued that the alleged jailbreak was merely a technique involving asking the model to read a specific codebase and fix software flaws—a standard capability widely available in rival models, including OpenAI's GPT-5.5.[2][6]

"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, warning that applying this standard across the industry would "essentially halt all new model deployments for all frontier model providers."[1][6]

The models at the center of the dispute represent the cutting edge of Anthropic's technology. Fable 5 was released to the general public with strict safety classifiers designed to block responses related to cybersecurity, biology, and chemistry.[4][5]

Mythos 5, built on the same underlying architecture, had those specific constraints lifted and was restricted to a vetted group of critical infrastructure operators and cyber defenders. Anthropic noted that before launch, both models underwent thousands of hours of "red-teaming" by the U.S. government and the U.K. AI Security Institute, neither of which found a universal vulnerability.[2][5][6]

The shutdown marks a historic shift in Washington's approach to AI regulation. For years, U.S. export controls have focused strictly on the physical hardware—specifically the advanced semiconductors and manufacturing tools required to train AI systems.[1]

Friday's directive represents the first time the federal government has used export controls to restrict access to the software and model weights themselves, effectively treating commercial AI access as a controlled munition.[1][4]

The move follows months of escalating tension between Anthropic and the federal government. Earlier this year, the Department of Defense labeled the company a "supply chain risk" after Anthropic refused to allow its models to be used for domestic surveillance or fully autonomous weapons systems.[1][5]

The political climate surrounding AI has also tightened. The directive comes just ten days after President Donald Trump signed an executive order establishing a framework for federal agencies to vet the national security risks of advanced AI systems prior to their public release.[3]

Defense officials publicly backed the Commerce Department's aggressive stance. Kirsten Davies, the Pentagon's chief information officer, posted on social media that national security must take precedence over commercial interests, stating, "Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation. America First. Always."[1]

The global fallout from the directive was immediate, sending shockwaves through the international tech community. Foreign developers and allied nations were suddenly confronted with the reality that the U.S. government can unilaterally sever access to foundational AI infrastructure without warning.[4]

This realization is already accelerating conversations around "AI sovereignty," as international companies and governments recognize the severe operational risks of relying on U.S.-hosted frontier models that can be shut down by an executive branch directive.[4]

Anthropic has stated it is working urgently with the government to resolve the dispute and restore access to Fable 5 and Mythos 5. In the meantime, the company's older models remain online, but the broader AI industry is now operating under the chilling precedent that a single perceived vulnerability could trigger a global shutdown.[1][2][6]