Behind Anthropic's Shutdown, a Relationship With Washington That Had Already Broken Down

The Fable shutdown is starting to look less like a sudden security scare and more like the end of a relationship that had been fraying for months. New reporting Monday traces the trigger to a breakdown in trust between Anthropic and the Trump administration, sitting on top of the jailbreak itself. One administration official put it bluntly: they screwed us. That is not how people describe a routine technical dispute.

This did not start last week. The friction goes back to February, when the administration moved to stop federal agencies from using Anthropic's tools, and the Pentagon labeled the company a supply chain risk. The root of that fight was Anthropic refusing the Pentagon unrestricted access, drawing lines around mass domestic surveillance and fully autonomous weapons. So by June, the two sides were already deep in a standoff.

Per Axios, the spark came when Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent to flag that Mythos and Fable could be jailbroken. The administration says it had also learned the Mythos model was made available to an entity in a foreign country with direct ties to the Chinese Communist Party, which sharpened everything. Officials say Anthropic failed to take a cyber executive order seriously, and its initial stance was, in one source's telling, that this is not a real issue and they would not act.

Anthropic tells it differently. The company says the government handed over only verbal evidence of a narrow, non-universal jailbreak, with no written specifics on the threat. It argues the bypass is minor, reproducible on other public models, and not a real hole in the model's safeguards. From Anthropic's seat, it was being asked to pull a flagship product on thin, undocumented grounds. Two sides, two versions, and a lot of bad blood underneath.

The reframe matters for how this ends. If the shutdown were purely technical, a fix and a sign-off would settle it. A relationship problem is harder. It explains why Anthropic flew senior staff to Washington over the weekend to rebuild trust before its IPO, beyond any technical fix. And it raises the uncomfortable question of how much of US AI policy now runs on personal rapport between founders and officials.

So the cleanest read of the Fable saga may be the least technical one. A safety dispute became a shutdown because the people involved had stopped trusting each other. Models can be patched in an afternoon. Relationships take longer. That is the part Anthropic is now trying to fix in person.