WASHINGTON, D.C. — In what is being characterized as the most aggressive federal intervention into the domestic artificial intelligence ecosystem to date, the Trump administration has issued a sweeping, unexpected export restriction that has forced top-tier AI developer Anthropic to pull its most advanced flagship models from the global marketplace.

The late-Friday directive from the U.S. Commerce Department caught Silicon Valley entirely by surprise, disrupting the momentum of a highly anticipated, newly minted trillion-dollar public market trade environment.

At the center of the dispute are Anthropic’s unreleased vanguard models, Fable 5 and Mythos 5. Citing acute national security vulnerabilities, federal officials ordered an immediate block on all foreign nationals accessing the systems—a mandate so restrictive it ultimately forced Anthropic to pull the products from the market entirely.

The Anatomy of the Restriction: Why Anthropic Pulled the Plug

The Commerce Department’s order specifically targeted access by foreign nationals, a definition that legally encompasses not only overseas users but also foreign citizens residing and working inside the United States.

Faced with the logistical impossibility of vetting the operational access of thousands of global developers, researchers, and even its own international staff members, Anthropic executives made the strategic choice to take both models offline worldwide while high-level negotiations with Washington continue.

The government’s sudden intervention hinges on a critical software vulnerability discovered within the models’ core architecture:

• The Guardrail Vulnerability: Federal cyber intelligence units flagged a specific structural exploit, colloquially known as a “jailbreak” in the source code.
• The Threat Vector: National security experts warn that this narrow flaw could allow adversarial actors to completely bypass Anthropic’s embedded safety guardrails, potentially weaponizing the AI’s advanced code generation to orchestrate sophisticated cyberattacks against critical infrastructure.

While Anthropic leadership has publicly downplayed the severity of the flaw—arguing the vulnerability is narrow and highly useful for “white hat” cyber defenders to build better protections—the Trump administration is refusing to compromise on digital frontier safety.