US imposes export controls on Anthropic after jailbreak and security issues, sparking DeFi AI shift

White House technology adviser David Sacks said the Trump administration imposed export controls on Anthropic after the company declined to patch a reported jailbreak in its Fable AI model. Sacks also described testing that found a way around guardrails intended to block the model’s use for cybersecurity ^[1].

Multiple reports say the administration moved to restrict international access to Anthropic’s models after a security alert tied to Amazon’s internal research about capabilities in Anthropic’s Fable 5. According to those accounts, officials required Anthropic to remediate or withdraw the model, and President Trump authorized comprehensive export restrictions that led Anthropic to terminate access to Fable 5 and Mythos for its user base ^{[2] [3]}.

Separately, Crypto Briefing reported that Chamath criticized Anthropic for allegedly “nerfing” models for frontier research users, arguing the approach could erode trust and complicate compliance for enterprise AI customers ^[4].

In the latest development, U.S. actions described by officials and media reports have resulted in export controls and access restrictions affecting Anthropic’s Fable and Mythos models, while industry figures have publicly raised concerns about Anthropic model handling for certain user groups ^{[1] [2] [4]}.