Anthropic shuts down access to AI models after US government ban on foreign nationals

Anthropic pulled the plug on its two newest AI models just three days after launching them. The reason: a US Commerce Department directive banning foreign nationals from using them, issued under broad national security concerns.

The company determined it couldn’t reliably enforce selective access, meaning it couldn’t just block non-US users while keeping the models live for everyone else. So it shut everything down.

What happened and why it matters

On June 9, Anthropic released Claude Fable 5 to the general public and Claude Mythos 5 to a restricted group of approved users. Three days later, on June 12, the Commerce Department issued a directive suspending access to both models for all foreign nationals.

Anthropic’s response was to disable the models entirely. The company concluded that selectively enforcing the ban, trying to verify the nationality of every user across every platform, was simply impractical.

Previous US export controls on AI technology focused almost exclusively on hardware, restricting the sale of advanced semiconductors to countries like China. This directive breaks new ground by targeting the AI models themselves, the software layer that runs on top of all that hardware.

The Commerce Department cited national security concerns but didn’t provide specific evidence or a detailed rationale for why Fable 5 and Mythos 5 warranted such drastic action. Among the concerns mentioned was the risk that users could jailbreak the models, bypassing built-in safety guardrails to extract dangerous capabilities.

Who gets hurt

The immediate fallout lands on enterprise customers, API users, and partnerships running through major cloud platforms. That includes AWS, Google Cloud, and Microsoft, all of which serve as distribution channels for Anthropic’s models.

Any company outside the US that had integrated Fable 5 or Mythos 5 into its workflows is now looking at a sudden and complete loss of access. No transition period, no workaround. The models are simply offline.

Mythos 5 was restricted to approved users from day one. None of that mattered when the government order came down.

The bigger picture for AI regulation

Extending export controls to AI model access represents a qualitative shift beyond hardware restrictions. Hardware restrictions are relatively straightforward to enforce. You can track who buys a chip. Enforcing restrictions on who gets to use a cloud-based AI model is a fundamentally different problem, which is exactly why Anthropic decided it was easier to just turn the lights off.

What this means for investors

For anyone with exposure to AI companies, whether through direct investment, cloud infrastructure plays, or enterprise software that depends on frontier models, this is a signal worth taking seriously.

Cloud providers that distribute these models face their own headaches. AWS, Google Cloud, and Microsoft all serve global customer bases. Having to enforce nationality-based access restrictions on specific AI models adds compliance complexity that none of them signed up for.

Investors should watch for two things going forward. First, whether this directive gets applied to other AI companies and models, or whether Fable 5 and Mythos 5 were singled out for specific reasons the government hasn’t disclosed. Second, whether allied nations push back through diplomatic channels, which could reshape how broadly these restrictions ultimately apply.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.