The US government ordered Anthropic to block access to its Claude AI model for all non-US nationals, treating the software as a restricted export item without new laws or extensive deliberation.

What happened

The US Commerce Department requested that Anthropic restrict access to its Claude models (referred to as Mythos and Fable 5 in the article) to US citizens only, following a report from Amazon (Anthropic's largest investor) to the government about potential misuse. Anthropic ultimately chose to restrict the models for everyone rather than attempt to filter by citizenship, as implementing citizenship-based access controls on globally deployed software is technically unfeasible.

Why it matters

This marks the first time the US has applied trade restrictions at the AI model layer rather than only at the chip level, signaling a shift toward treating advanced AI systems as geopolitical weapons. The decision was made without new legislation or broad public deliberation, and it demonstrates how one nation can unilaterally deny access to widely used software to the rest of the world. This is likely to reshape how other nations and companies approach AI development and may accelerate efforts by other countries to build independent AI capabilities.

What to watch

The article suggests second- and third-order consequences will follow, including a decline in public research on LLMs, nations hoarding breakthroughs, and for-profit labs yielding more to government regulation. The author also notes that frontier models with superior capabilities will continue to arrive via data centers and APIs rather than local devices, making AI access and geopolitical control over model availability an ongoing point of international tension.