White House Forces Shutdown of Anthropic Fable After Amazon Alert

Anthropic Fable pulled after Amazon‑reported jailbreak

The release of the Anthropic Fable model on June 9 was halted within days when Amazon warned the White House of a security breach that could threaten national interests. Inside sources say Amazon delivered a report Thursday night showing how its team accessed portions of Anthropic’s Mythos engine, prompting senior officials to act swiftly.

White House response

By Friday evening, a formal export‑control notice was sent to Anthropic, citing a ↓ 90 minutes window to disable the model. The letter, first reported by Reuters, placed Fable and the underlying Mythos under the same restrictions that already bar foreign adversaries.

“The threat level was presented as immediate, yet the technical details were vague,” a senior official told Bloomberg.

Anthropic’s CEO Dario Amodei argued the jailbreak was “relatively simple” and did not expose a flaw in the model’s safety layers. He warned that the abrupt shutdown would affect foreign‑born staff who rely on the system for daily work.

Industry observers note the move creates a de‑facto licensing regime: any AI system approaching Mythos‑level capability may require prior clearance. The episode underscores how quickly policy can overtake development in a sector racing ahead of regulation.