US blocks Anthropic AI model for non-American users

Anthropic has closed its AI model Fable, a slimmed-down variant of the more powerful Mythos model, to users outside the United States. Only American citizens can now access the model. European companies and developers who were working with Fable have consequently lost access.

It is one of the first times a major American AI company has drawn such a sharp geographic boundary around a specific model. Until now, most American AI models were available worldwide, with the exception of a limited number of countries already subject to sanctions. An exclusion targeting Europe specifically is new.

The exact reason Anthropic has given has not been made fully public, but the measure fits within a broader trend of the US government seeking greater control over the export of advanced technology.

What Anthropic has blocked

Fable is a version of Anthropic's Mythos model made more accessible for broader use. It is not an experimental or early-stage prototype, but a model that was already actively used by developers and companies outside the US. The restriction applies to all non-American users, including those in the EU.

Anthropic has implemented this measure itself. That suggests the company is acting on the basis of regulation or contractual obligations, possibly agreements with the US government concerning the deployment of advanced AI systems. Anthropic has not issued an extensive public statement on the matter, according to available reporting.

Export controls expanding into AI

The US has been gradually tightening export controls on technology with strategic or military value for years, covering chips and software with military or strategic applications. AI models have largely escaped that regulatory framework so far, but Anthropic's move suggests that is beginning to change.

In recent years, Washington has tightened the reins on advanced technology through legislation such as the Export Administration Regulations and sector-specific restrictions. If AI models are explicitly brought under comparable frameworks, it will have significant consequences for the way European companies and governments can access American AI infrastructure.

For now, the measure around Fable is an isolated case, but it does set a precedent. If other American AI providers take similar steps, the accessibility of the global AI offering will change structurally.

Implications for European users and businesses

For European developers and companies using Fable, the impact is immediate. They must find alternatives, which involves time, integration work, and potentially higher costs. Companies that have built their products or services on a specific model are particularly vulnerable when access suddenly disappears.

More broadly, this underscores how dependent European AI applications are on American providers. The majority of leading large language models currently come from the US. Access to those models is generally taken for granted, but the situation surrounding Fable shows that such access can change at any moment.

What this means for the European AI scene

For policymakers and investors in Europe, this situation confirms that building domestic model capacity has strategic value. The EU has already taken steps towards European computing power and model development through the AI Office and initiatives such as EuroHPC, but the gap with American providers remains substantial.

Current access to American models says nothing about the availability of those models in the future. For Dutch and European founders building scalable AI products, this makes the choice of underlying infrastructure a strategic consideration, not merely a technical one. Initiatives focused on open or European-controlled models gain additional weight as a result in the investment decisions being made by investors and policymakers.