How Amazon and the White House ended Anthropic's Fable
TL;DR
Anthropic released Fable 5 on June 9 as a general-use version of the more powerful Mythos 5. Within days, access was gone after an Amazon security report reportedly triggered pressure from the White House. Amazon told US officials it had used prompts to reach parts of Mythos 5 that could pose national security risks. The tension is obvious: Amazon is also a major investor in Anthropic.
Nauti's Take
Amazon is wearing two hats here, and that costs trust: investor in Anthropic, alarm bell inside Washington. If the report mainly showed that a model can answer defender-style security questions, the escalation looks like politics wearing a security badge.
Anthropic still has work to do: if it opens access to a Mythos-level model, it needs clear rules for allowed capabilities, logging, and pause triggers before launch. The industry needs thresholds, or the loudest call decides who gets access.
Briefingshow
The case shows how unstable access to frontier models has become: one security report, several company calls, and political pressure can apparently take a model offline overnight. For teams building AI into products or security workflows, that is a new platform risk. The line between legitimate cyber defense and banned offensive capability remains blurry.
Sources