Anthropic’s Own Words Triggered US Export Ban on AI Models

Anthropic’s public alarm bells rang louder than anyone else’s in AI last year. The company hammered home risks and safety concerns far more than its rival OpenAI. The effect was unmistakable: Washington slapped an export ban on Anthropic’s latest models, Mythos and Fable, in June 2026.

In 2026, Anthropic mentioned risk, regulation, or restrictions five times every 1,000 words it published. OpenAI, by comparison, barely hit 0.6 words per 1,000. Anthropic’s communications featured the word “risk” 336 times. OpenAI used it just 30 times. “Safeguard” appeared 121 times at Anthropic, only 33 times at OpenAI. “Vulnerability” was said 128 times versus OpenAI’s 10.

Anthropic launched Mythos in April 2026, presenting it as a tool to uncover cybersecurity weaknesses and vulnerabilities. The company emphasized the model’s risks to critical infrastructure, the financial sector, and national security. Mythos’ rollout was tightly controlled, initially limited to select U.S. organizations for safety.

The U.S. Department of Commerce forced a freeze on foreign nationals accessing Mythos and Fable on June 12, 2026. Anthropic promptly suspended access to these models in response. The ban came after months of friction between Anthropic and government agencies, including a Pentagon warning in February 2026 that flagged Anthropic as a supply-chain risk to national security. Anthropic is now challenging that designation in court.

Behind the scenes, Anthropic had worked with officials on a controlled Mythos rollout. Publicly, they criticized regulators for dragging their feet on AI oversight. The company’s messaging was blunt and relentless. Critics like OpenAI CEO Sam Altman called it “incredible marketing” to announce, “We have built a bomb. We are about to drop it on your head.” Meta’s former AI chief Yann LeCun branded it “fear-mongering” and “ridiculous.”

Anthropic’s tone softened since 2023, halving its use of risk-related language. But in 2026, the company doubled down. The warnings and self-imposed restrictions painted Mythos and Fable as powerful but dangerous tools. This narrative likely spooked regulators. It made the U.S. government feel compelled to clamp down before these models fell into foreign hands.

Mythos became one of the most talked-about AI models of 2026. Interest peaked after its unveiling and the export ban. The episode exposed a paradox: Anthropic’s stark warnings about AI risks helped trigger the very restrictions they hoped to avoid. By talking up safety concerns so loudly, they may have boxed themselves in.

The U.S. government and Anthropic have a history of clashes over AI’s role in surveillance and autonomous weapons. Now, with Mythos and Fable caught in a regulatory vise, the company faces a tough balancing act. They must prove their models are safe without sounding like they’re holding a ticking time bomb.