Artificial intelligence company Anthropic has taken its two newest and most advanced AI models offline after receiving a directive from the Trump administration. The step is being described as one of the most sweeping directives yet aimed at an AI company, and it marks another significant test of the increasingly strained relationship between big technology firms and the United States government.
According to Anthropic, the directive ordered the company to block foreign nationals, both inside and outside the United States, from accessing two of its most advanced and recently released models, identified in the broadcast as Mythos 5 and Fable 5. The administration argued that the models represented a national security concern, although the company said the government did not specify in its letter exactly why that was the case.
To comply with the order, Anthropic disabled both models entirely, making them unavailable not only to foreign nationals but to all users. The company said it stood by the safeguards it already had in place for the models, even as it moved to follow the government's instruction. The decision effectively pulled two state-of-the-art systems out of public reach in a single step.
Anthropic warned that the implications of the directive reached far beyond its own products. The company argued that the standard being imposed would essentially halt all new AI models across the industry, suggesting that other developers could face the same obstacles if the approach were applied broadly. That framing positioned the dispute as a potential turning point for how advanced AI is released in the country.
The move follows an executive order that the president signed earlier this month asking technology companies to give the administration access to their AI models before releasing them to the public. The directive to Anthropic is being read as part of that wider push, in which the government is seeking far greater oversight of the most powerful systems before they reach users.
Separately, Anthropic has sued the Pentagon after being labeled a supply chain risk, even as it continues to work with the United States government in other areas. Taken together, the offline models, the executive order and the lawsuit point to escalating friction between the AI industry and Washington, with companies and officials increasingly at odds over national security, access and control of the technology.