Anthropic announced Friday that the U.S. Commerce Department ordered it to suspend access to its Mythos and Fable AI models for all foreign nationals, citing national security risks. The San Francisco-based company said it must abruptly disable the systems to ensure compliance, affecting both external customers and its own international employees.

The restrictions target Anthropic's most advanced AI systems, which the company acknowledged could become powerful tools for hackers targeting computer networks. Since unveiling Claude Mythos in April, the company shared limited access with roughly 40 critical infrastructure organizations to help patch vulnerabilities before malicious actors could exploit them.

Former Trump administration officials called the decision baffling and out of step with recent policy moves, including allowing advanced AI chips to be sold to China. The order reflects growing tension between innovation and security oversight as the administration shifts toward more regulation of emerging AI technologies.

This move follows OpenAI's similar restriction of GPT-Cyber 5.4, though that company initially shared its technology with hundreds of experts before broader release. The unprecedented action signals Washington's increasing willingness to intervene in AI development over security concerns.