The U.S. government ordered Anthropic to suspend access to its latest Claude models, marking the start of a new era where executive branch decisions on AI releases are made with minimal technical expertise.

What happened

On Friday after markets closed, the U.S. executive branch reached out to Anthropic and forced the company to suspend access to its Claude 5 Mythos/Fable models for any foreign national or user abroad. The White House was alerted to cybersecurity concerns about a jailbreak of the Fable model by Amazon, Anthropic's largest financial and technology partner.

Why it matters

This signals a shift from the ChatGPT-era governance model (where AI was seen mainly as a question-answering tool) to what the article calls the AGI era, where governments feel they must act quickly to control increasingly capable AI agents that complement human workers. The episode reflects deep uncertainty and political pressure: model releases are now being judged by an executive branch with minimal technical talent, using politically influenced technical assessments rather than consistent technical standards.

What to watch

The article describes this as just the starting gun—a new normal where sovereign governments worldwide (Europe, the Middle East, China) will assert control over frontier AI, and similar aggressive actions against open-source models are likely to follow, though the timeline remains unclear (potentially 3 months to 2 years).