On Friday afternoon, the U.S. Commerce Department sent Anthropic an enforcement letter invoking an obscure export control directive. The letter effectively forced the company to pull Fable 5 and Mythos 5—its two latest AI models—offline to all customers, both domestic and foreign. The directive banned non-Americans, including Anthropic employees, from accessing the models, citing an unspecified national security concern.
Anthropus said it believes the ban relates to a bypass of the models' guardrails but acknowledged the letter provided no specific technical details. The letter itself has not been made public. In response, the company shut down both models to ensure compliance, a swift and unilateral action that did not appear to require court approval.
According to reporting cited in the article, the enforcement action may stem from a technical paper written by Amazon security researchers describing an alleged guardrail bypass in Fable 5. Katie Moussouris, a cybersecurity veteran and founder of Luta Security, reviewed the paper and released a blog post examining the findings. She concluded that the behavior described—a difference between asking a model to "review code for security issues" versus asking it to "fix this code"—should not have triggered an export control designation.
"The behavior described in the paper cannot meaningfully be fixed, and any attempt would only weaken the model for defense," Moussouris said, criticizing the export control directive as "hasty, heavy-handed, and misguided." She and dozens of other top security researchers have since called on the Trump administration to revoke the order, arguing that removing advanced cybersecurity capabilities from U.S. network defenders is "dangerous."
The Trump administration has not publicly confirmed why it invoked the export control directive. Axios reported that "personality differences" between Anthropic and the Trump administration may have driven the decision, rather than a technical issue with the products. Justin Hendrix, editor of Tech Policy Press, said the move is "likely to raise alarms in foreign capitals about the reliability of American AI for critical applications" and that it creates a climate where "senior officials are picking favorites based on personal and political factors."
The article notes that past administrations have made sweeping decisions on export law based on incomplete knowledge. During the 2010s, government language covering cybersecurity tools was so broad that it nearly outlawed legitimate security and vulnerability research. However, the Trump administration's directive appears to operate differently: it forced a company to pull products offline without a public explanation or court involvement.
The precedent is significant. The government has demonstrated it can unilaterally shut down AI products without judicial review, setting the stage for future interventions in the tech industry. As the article states: "This time the government took issue with Anthropic; tomorrow it could be with anyone else."