The Sovereign AI Trap: Why Infrastructure Isn't Control

As organizations rush to adopt "sovereign AI" through domestic compute capacity and homegrown models, they may be inadvertently narrowing their own future options.

According to reporting from The Register, Mike Bracken—founder of GOV.UK, former UK Government Chief Data Officer, and co-founder of consultancy Public Digital—warns that institutions are risking a deeper reliance on a small group of AI providers. Bracken argues that true sovereignty is not defined by owning the underlying infrastructure or keeping technology within national borders, but by whether an organization retains the ability to make its own decisions when circumstances change.

In his book, *Digital Sovereignty: The Power to Decide*, Bracken notes that institutions often lose control not during a crisis, but through a series of "reasonable" decisions that gradually build dependencies. For commerce operators and other institutions, the risk is that AI services become so deeply embedded in workflows and infrastructure that switching suppliers or strategies later becomes prohibitively difficult.

While the UK government continues to invest in domestic computing infrastructure to reduce reliance on overseas providers, Bracken suggests this misses a fundamental point. He asserts that calling AI "sovereign" is meaningless if an organization remains locked into the specific supplier that provided the system, ultimately constraining their freedom to act.