Anthropic Takes AI Models Offline After Trump Directive

U.S. Government Orders Removal of Advanced AI Models

Anthropic has taken its latest artificial intelligence models offline following a Friday afternoon directive from the Trump administration aimed at preventing foreign nationals from accessing the technology. The company removed both Fable 5, which it had just released widely this week, and Mythos 5, a more advanced model with tightly restricted access.

The move represents the most substantial U.S. government action to date in restricting access to cutting-edge AI systems, according to CBC News, which first reported the development.

Why It Matters

This action signals a new era of direct government intervention in AI deployment, moving beyond voluntary frameworks to immediate enforcement. For enterprise technology leaders, it creates uncertainty around access to frontier models and suggests that national security reviews may become standard gatekeepers for advanced AI capabilities, potentially disrupting deployment timelines and international collaboration.

Company Disputes Process

In its public statement, Anthropic expressed disagreement with how the administration handled the situation. The company said it received no specification of the national security concerns prompting the action and called for a "statutory process that is transparent, fair, clear, and grounded in technical facts."

Anthropic characterized the situation as a "misunderstanding" and indicated it hopes to restore access to the models quickly. The Commerce Department has not yet commented on the directive.

Context of Recent Executive Order

The directive arrives ten days after President Donald Trump signed an executive order establishing a framework for federal vetting of advanced AI systems. That order allows the government to review national security risks for up to one month before public release of the most sophisticated models, though participation was described as voluntary.

When Anthropic introduced Mythos in April, the company acknowledged the model's capability to identify software vulnerabilities at scale could pose serious risks to economies, public safety, and national security if misused. Mythos represents a more powerful system than Fable, which itself is described as a limited version of the flagship model.

Broader Implications for AI Governance

Bank of Canada governor Tiff Macklem addressed Mythos during International Monetary Fund spring meetings in April, noting substantial discussion about the model among global financial leaders. Macklem emphasized that advanced AI models with the ability to expose and exploit vulnerabilities place a premium on robust cybersecurity protections.

He also stressed that Mythos should not be viewed as an isolated event. The rapid evolution of AI technology requires firms, regulators, and policymakers to develop comprehensive plans for managing these capabilities as they continue to advance.

The details were first reported by CBC News.