Research Labs

AI Sovereignty Lab

Who controls the AI your country or organization depends on — and what happens if access is taken away?

Introduction

As our business processes and public services become increasingly dependent on AI, we are exposing ourselves to new kinds of risk. The essential building blocks of AI — chips, compute, cloud infrastructure, foundation models and the data they are trained on — are concentrated in the hands of a small number of companies and countries.

Control over these AI resources has become a political instrument. The US government, for example, has shown its willingness to restrict access to certain AI resources for other nations, and for institutions it disfavors. Access that was taken for granted yesterday can be limited tomorrow by a policy decision made elsewhere.

At the national level, countries must strive for an appropriate degree of independence. Few, if any, can realistically own the entire AI stack, so each must determine which layers matter most for its security, economy and public services — from energy and compute to models, data, applications and talent — and where partnerships or open alternatives are acceptable.

At the level of organizations, CEOs and boards need to look strategically at their exposure to AI infrastructure dependencies, and mitigate risks that are unacceptable for the business: vendor lock-in, sudden changes in pricing or terms of service, withdrawal of services, and the jurisdiction their data and models fall under.

Topics addressed in this lab

Recent and ongoing work

The lab's first publications are in preparation and will be listed here. Interested in contributing to this lab? Get in touch.

Our labs