Austria Pushes EU to Court Anthropic as US Tightens AI Access

Austria has formally urged the European Union to consider bringing Anthropic into European territory, just weeks after the United States imposed sweeping restrictions on foreign access to the company's most powerful AI systems.

State Secretary for Digitalization Alexander Pröll authored the proposal in a letter addressed to European Commission Executive Vice President Henna Virkkunen. While making the case for action, Pröll openly acknowledged that the practical details of such a move remain unclear.

The core of Pröll's argument rests on urgency. He warned that unless Europe moves quickly, the continent risks being permanently sidelined from cutting-edge AI development. His letter floated the idea of "the strategic establishment and participation of Anthropic within the European Union," dangling a package of incentives including legal stability, access to fresh capital, and full integration into the EU's single market. No specific funding commitments, construction timelines, or implementation frameworks were included in the proposal.

The backdrop to Austria's push is a dramatic policy shift out of Washington. On June 12, the US Commerce Department issued an export control directive targeting Anthropic's two most advanced AI models — Claude Fable 5 and Claude Mythos 5. The order barred all foreign nationals from accessing the systems, including non-citizen employees working at Anthropic itself. The directive cited national security as justification.

The timing proved particularly disruptive. Both models had only recently launched when the restrictions took effect. Since Anthropic had no mechanism to verify users' nationalities at scale, the company took both models offline globally. Claude Opus 4.8 remained available as a fallback.

The security concerns stemmed partly from an incident involving Amazon, Anthropic's largest financial backer. Amazon researchers reportedly extracted restricted cyberattack-related content from Claude Mythos 5. CEO Dario Amodei characterized the breach as a narrow workaround rather than a full jailbreak, though the episode drew attention to the model's demonstrated ability to probe secured government infrastructure.

By June 26, US authorities had partially walked back the restrictions, restoring access to Claude Mythos 5 for more than 100 vetted American institutions. Claude Fable 5, however, remains under the export block.

For all the political appeal of Austria's proposal, the structural obstacles are considerable. Anthropic is currently financing a $50 billion data center expansion across Texas and New York. Amazon has committed $13 billion to the company and serves as its primary training infrastructure partner. In exchange, Anthropic has pledged to spend upward of $100 billion on Amazon's cloud services over the coming decade.

Anthropichas also projected that US AI infrastructure will require approximately 50 gigawatts of additional power capacity by 2028. Europe's position on all the key inputs — compute, energy, and capital — lags significantly behind. The EU's Chips Act set an ambitious target of capturing 20% of global semiconductor production by 2030, up from under 10% today. The bloc's own internal forecasts, however, place the realistic figure at just 11.7%, and EU auditors have described the original goal as highly improbable.

The European Commission now faces a question that is as much political as it is logistical: does Austria's appeal represent a viable strategic opportunity, or is it primarily a statement of intent in the escalating global contest over AI supremacy? Europe holds genuine regulatory influence through its AI Act, but influence over rules is a different matter from owning the infrastructure that powers frontier models.

Whether Brussels treats this as an actionable proposal or a diplomatic signal is likely to become apparent in the weeks ahead.