Trump Admin Uses Obscure Export Rules to Block Anthropic AI Models
First time US export controls used to restrict access to AI models via cloud.
The Trump administration has ordered Anthropic to block access to its newest AI models — Fable 5 and Mythos 5 — for all foreign nationals, including users within the US and even the company's own employees. The government cited export control rules under national security authorities, but has not publicly explained the legal basis. This marks, according to experts, the first time US export controls have been used to restrict access to an AI model hosted remotely on a cloud service, rather than physical goods or downloadable software.
Anthropic complied by cutting access for everyone, including US users, to avoid violating the order. The move exposes a critical gap in current export control frameworks — remote access to AI models via APIs or chatbots doesn't neatly fit existing categories. Georgetown's Hanna Dohmen calls it an "open question" whether the order stretches rules beyond their intended scope. The episode raises alarm for other labs like OpenAI, Google, and Meta, as the government hasn't clarified what capabilities or safeguards triggered the action.
- Trump admin used export control orders to block foreign nationals from accessing Anthropic's Fable 5 and Mythos 5 models.
- The order is unprecedented — export controls have never before been applied to remote AI model access via cloud.
- Anthropic had to block all users worldwide to comply; legal basis remains unexplained, creating industry-wide uncertainty.
Why It Matters
Sets a precedent for opaque AI governance that could disrupt access to frontier models globally.