The US government on Friday imposed export controls on Anthropic’s most powerful AI model, Claude Mythos. It was a stark reminder to Europe of why relying on tech it does not own or control threatens to undermine its sovereignty.
AI is set to be the defining technology of the decade and likely beyond. If Europe wants a say in its development and application, it needs to build an ecosystem which helps its own tech companies compete internationally. Better functioning capital markets are part of the solution.
After the US government intervention, Anthropic cut off public access, including inside the US, to its new Fable model, a version of Mythos with extra safeguards. Although the export controls technically only applied to foreign nationals, Anthropic said it had to suspend all access “to ensure compliance”.
Fable was considered by many to be the most capable publicly available AI model.
The US government’s increasingly interventionist approach to AI regulation should worry countries that are integrating US AI models into their economies. They will be forced to accept what the US decides, unless they can obtain enough leverage to make their views matter.
Leverage how?
The question for Europe is how it can obtain leverage. There is no silver bullet.
It is fanciful to think that European firms can close the gap to the US AI labs any time soon. The answer is more likely to be a combination of having influential tech companies and more importantly, developed digital infrastructure.
Availability of computational capability is a key constraint on the pace of AI development. Europe needs to build the data centres to feed at least its own demands.
However, at the moment, Europe’s digital infrastructure lags that of the US. There are structural differences, like the price of energy and constraints on space, that make it difficult to build data centres on the scale that the US has managed.
But another problem is the structure of European capital markets. Europe’s reliance on bank debt means bank appetite for data centre lending threatens to constrain how quickly infrastructure can be built.
It is also more challenging for early stage companies in Europe to raise money compared to their US peers. That makes scaling up tech firms more of a challenge.
In the US, securitization has been crucial in financing data centre development. The data centre asset-backed securities market is starting to get going in Europe, but there are extra obstacles. European securitization markets are not as deep as the US market and regulatory fragmentation makes it more difficult to put together enough collateral to back a securitization.
Other established digital infrastructure asset classes in the US, like fibre ABS and tower ABS, are yet to emerge in Europe.
Both the EU and the UK are in the process of reforming their securitization rules. Both are aiming to expand the market to strengthen their economies, but there are doubts about whether they are taking the steps necessary to achieve that ambition.
Implementation of any reforms will take time. The EU is set to reach a final text before the end of this year, after a process that began at the end of 2024. But even then, more time will be needed for implementation.
All the while, AI keeps developing at breakneck pace. If Europe wants a seat at the table, it needs to make its voice matter.