Anthropic Refused To Patch Claude Fable's Jailbreak, So The US Banned It, David Sacks Says

David Sacks, a White House technology adviser, said the Trump administration imposed export controls on Anthropic after it declined to patch a reported jailbreak in its Fable AI model.

Sacks Details Anthropic Order

In a post on X on Saturday, Sacks wrote that a trusted partner testing Fable had found a way around guardrails meant to block its use for cybersecurity. The administration then asked Anthropic chief Dario Amodei to fix the issue or pull the model.

Amodei refused, according to Sacks.

The dispute followed Anthropic's Jun. 9 release of Fable 5, a version of its more powerful Mythos 5 system fitted with safety guardrails. Commerce Secretary Howard Lutnick later barred foreign nationals from both models, and Anthropic disabled them worldwide.

Also Read: Bitcoin Bulls Eye $67K After Trump Says Hormuz Will Open To All

Anthropic Disputes Jailbreak Severity

Anthropic pushed back in a blog post. The company said it disagreed that a narrow jailbreak should force the recall of a model used by millions, and argued that rival systems could be exploited the same way.

Sacks said the move was reluctant, and that officials hoped Anthropic would fix the flaw so Fable could return. He also denied any link to earlier friction. In March, the Pentagon labeled Anthropic a supply chain risk, a designation the company is now contesting in federal court.

Read Next: Index Rules Turn SpaceX's $2T Debut Into A Market Stress Test