For the first time in its history, the U.S. Department of Commerce invoked authority from the 2018 export-control reform law to order a blanket export ban on Anthropic’s latest AI models. In a Friday letter, Commerce Secretary Howard Lutnick instructed Dario Amodei, CEO of Anthropic, to immediately halt exports of the Mythos and Fable models to all destination countries and to all foreign nationals worldwide.
The department justified the action on national security grounds, saying the models could be exploited by the military intelligence services of China, Russia, or other states that pose national-security risks. The government stated it had identified a technique that could bypass Fable 5’s cybersecurity protections, potentially making the model useful for discovering software vulnerabilities.
Anthropic responded in a Friday blog post, saying the technique would only reveal negligible security gaps that are also easily found in other freely available models. The Commerce Department relied on the 2018 export-control reform statute, which authorizes controls on newly emerging technologies deemed critical to national security. Experts noted that this is the first time the department has used that authority.
The letter warns that violations of the restrictions would carry immediate criminal and civil penalties. Analysts, however, flagged a substantial legal question: modern AI models are often made available via remote, cloud-based access rather than by traditional “export,” and current export-control rules do not clearly cover that mode of access. That uncertainty raises doubts about the department’s ability to enforce the ban.
The move deepens existing tensions between Anthropic and the Trump administration. Earlier this year Anthropic refused requests to allow its models to be used by the U.S. military for domestic surveillance and fully autonomous weapon systems; the administration responded by placing the company on a national-security blacklist. Anthropic had previously highlighted potential misuse risks in its Mythos model and delayed broader rollout for that reason.
According to the company, the public release of Fable 5—demonstrated on June 9—was tested in cooperation with the government and made available with the administration’s approval. On Sunday more than eighty cybersecurity leaders and industry experts, including staff from Nvidia and Adobe, sent an open letter to Howard Lutnick and National Cyber Director Sean Cairncross urging immediate lifting of the restrictions on Anthropic.
The decision affects Anthropic’s near-term commercial plans—including its confidential filing related to an impending public offering—and may set a precedent for regulating emerging AI technologies that the government views as national-security risks.
Practical implications
- The Commerce Department says the measure prevents the models from being repurposed for foreign military or intelligence use.
- Anthropic and its supporters argue the technical risks are limited or comparable to those of other publicly available models.
- Enforcement could trigger legal challenges, particularly over whether cloud-based access falls within existing export-control rules.
These developments are likely to reverberate across industry and policy circles as the debate over balancing security and innovation in AI continues.
