Andy Jassy reportedly told Treasury Secretary Scott Bessent that Amazon researchers used Claude Fable 5 to extract information useful for cyberattacks. That conversation, reported by The Wall Street Journal, apparently triggered the export control order that pulled Fable 5 and Mythos 5 from global access Friday.
The Commerce Department suspended all access to Fable 5 and Mythos 5 by any foreign national, including Anthropic’s own employees. David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, said the administration gave Anthropic CEO Dario Amodei an explicit choice: fix the jailbreak or de-deploy. Dario refused.
Anthropic pushed back hard. The company called the jailbreak “narrow and non-universal” — describing it as asking Fable 5 to read a specific codebase and fix software flaws. Anthropic also argued the same technique works on OpenAI’s GPT-5.5, which faces no similar export controls. The company is complying while contesting the order.
Here’s the structural problem: Amazon is Anthropic’s largest outside investor. An investor flagging its portfolio company to federal regulators isn’t how the AI investment playbook is supposed to go. It’s a reminder that in a national security frame, investor relationships don’t protect you. They can be the mechanism that exposes you.
For operators building on Claude, the practical read is this: the fastest path to a model cut-off isn’t a compliance failure on your end. It’s a conversation between your vendor’s biggest backer and a cabinet secretary. Add provider diversification to your vendor-risk checklist now.
Anthropic says it’s working to restore access. No timeline given.
— Nathan Zakhary