Anthropic received a letter ordering the immediate suspension of Claude Fable 5. The directive cited national security authorities and applied to "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." Anthropic complied within hours and killed the model for every customer worldwide, since it has no practical way to filter for nationality in real time. This comes to be the first government-forced takedown of a commercially deployed frontier AI model.
Fable 5 had launched only three days earlier. The trigger was reportedly a claim from another company that it had jailbroken Mythos 5, unlocking its cybersecurity capabilities in one specific scenario. Anthropic pushed back hard: it described the jailbreak as narrow and non-universal, and noted that the letter it received didn't even specify the exact security concern. The company called the situation a misunderstanding and said it is working to restore access, but one day later, nothing has happened yet.
The directive doesn't target adversary nations or specific use cases, but foreign nationals, including people sitting inside Anthropic's own offices. The practical message to every non-American working at a US AI company is clear: your access to the tools at the center of your job can be revoked overnight by executive action, regardless of your contract or contributions. The AI industry runs on global talent, and this order treats that talent's professional access as a foreign policy variable.
There is also a precedent that goes well beyond Anthropic. The US government has now established that it can treat proprietary AI models as export-controlled assets, the same regulatory category as advanced semiconductors or military hardware. Some AI figures publicly accused Anthropic of "regulatory capture based on fear-mongering," and more than one AI policy expert has said it's genuinely hard to tell whether this is legitimate national security policy or political pressure dressed in regulatory clothing.
And then there is the thing Anthropic walked into with its own messaging: the company built its identity around the idea that it was developing something really dangerous, capable enough to find security flaws at scale, powerful enough to require restricted access through Project Glasswing. That framing served it well for fundraising and policy credibility, but the government just used that exact framing as the justification for pulling the plug. In practice, Anthropic's safety communications handed regulators the vocabulary to act, and the argument for caution became the argument for shutdown.