Amazon flagged Claude Fable 5 — now the US has banned it

amazon ceo andy jassy reportedly alerted treasury secretary scott bessent and other officials to critical vulnerabilities in anthropic's claude fable 5, specifically that researchers used it to generate information for potential cyberattacks. this disclosure triggered the us government to impose export control bans on both fable 5 and mythos 5 models on friday, a rare move targeting specific ai weights. white house ai czar david sacks confirmed that a trusted partner reported the jailbreak and claimed the administration asked anthropic to fix the flaw or de-deploy, but ceo dario amodei refused. as amazon is a major anthropic investor, this event signals a fracturing point where private sector security concerns are directly driving immediate federal intervention in model deployment. the incident suggests that future ai governance may rely on adversarial reporting from corporate partners rather than standardized safety frameworks.