On a Friday evening in New York, Anthropic received a letter from the United States government. The message was brief and alarming: two of the company's newest AI models — Fable 5, widely available to the public, and Mythos 5, offered to existing subscribers — had to be taken offline immediately, citing unspecified "national security concerns." No supporting report was made public, no specific vulnerabilities were identified. Anthropic, which employs many foreign nationals, concluded that it had no practical way to comply with the order's demand to block all foreign users without simply withdrawing the models entirely.

The episode appears to have originated from inside one of Anthropic's closest business partners. Researchers at Amazon allegedly discovered a method to bypass Fable 5's safety guardrails, and Amazon CEO Andy Jassy is reported to have raised those concerns directly with the White House. What followed, in the words of TechCrunch's Sean O'Kane, moved "really fast, especially for a Friday afternoon into a weekend" — a speed that struck some observers as unusual, given that the administration was simultaneously engaged in sensitive diplomatic negotiations elsewhere. The order has not been welcomed by every corner of the security community.

A group of leading cybersecurity experts has signed an open letter calling on the Trump administration to revoke it, arguing that removing advanced AI-driven cybersecurity tools from American network defenders creates new dangers rather than eliminating old ones. Their concern is pointed: the models being restricted were also being used as shields, not just weapons.