Amazon CEO reportedly flagged security risks in Anthropic's AI models to U.S. officials, triggering a government export ban on two models.

What happened

Amazon CEO Andy Jassy told Treasury Secretary Scott Bessent and other government officials that Amazon researchers had used Anthropic's Claude Fable 5 model to obtain information that could be used in cyberattacks. The government subsequently imposed an export control ban on the Fable 5 and Mythos 5 models, and Anthropic cut off worldwide access to both. David Sacks, Trump's AI czar, stated that a trusted partner disclosed a 'jailbreak' (a method to bypass safety guardrails) in the model, and that Anthropic CEO Dario Amodei refused an administration request to either fix it or remove the model from service.

Why it matters

Amazon is a major investor in Anthropic, so its security concerns carry weight with policymakers. When a trusted technology company flags potential misuse of an AI model to the Treasury and other officials, the government can act quickly to restrict that technology—as happened here. This shows how security concerns raised by major cloud and AI players can directly shape which models remain available globally and which do not.

What to watch

AWS (Amazon's cloud business) has been affected by the model cutoff, meaning some customers may lose access to capabilities they were relying on. The incident also underscores the tension between companies like Anthropic (which rejected the government's initial request) and government officials over who controls the deployment of potentially risky AI models.