Read, Watch or Listen

Washington, United States – The U.S. government has imposed emergency export controls on Anthropic’s new consumer-facing AI model, Fable, following the discovery of a critical security vulnerability that regulators say could undermine its built-in safeguards. The action came after Amazon Chief Executive Andy Jassy alerted senior White House officials to findings from Amazon’s internal research team, which had uncovered a sophisticated jailbreak technique capable of bypassing Fable’s safety guardrails and enabling the generation of restricted, harmful or otherwise prohibited content. Federal regulators moved with unusual speed, giving Anthropic a 90-minute window to comply with a formal order that barred access for foreign nationals, shut down the model’s public-facing interfaces and required a comprehensive security audit to verify that its protections function as intended. Washington’s directive targets Fable, a consumer-focused, scalable version of Anthropic’s more powerful Mythos model, and effectively halts its planned market deployment while agencies including the Department of Commerce and the National Security Council review its technical specifications. Anthropic has publicly disputed the severity of the alleged vulnerability but says it is working closely with federal authorities to address their concerns and restore service, even as the White House maintains the restrictive controls on national security grounds. The move has rattled the broader technology sector, driving volatility in major AI stocks and prompting analysts to question whether this case signals a new standard of pre-emptive government oversight for frontier AI systems. It also raises wider questions about the future of AI exports and access, as the foreign-national restrictions highlight potential complications for U.S. firms seeking global reach and suggest a more fragmented, region-specific landscape for advanced AI deployment.

Prepared by Lauren Mitchell and reviewed by editorial team.