Europe’s billions in AI investment are creating opportunities for more than model builders and data centre operators. A new category of entrepreneurs is emerging around a less visible problem: how much control businesses actually have over the technology supporting their AI.

While companies race to deploy AI, many are doing so on infrastructure they cannot fully inspect, move or control.

Their data, models, software and computing increasingly depend on a small number of technology providers - which carries multiple risks.

A European Commission study published in August 2026 identified limited and geographically concentrated computing capacity as problematic. It also detailed other challenges difficult to ignore: dependence on non European cloud and AI suppliers and the potential risk of laws from third countries being applied.

As Dario Maisto, Principal Analyst at Forrester, points out, organizations are increasingly worried asking whether they can “ control how AI systems are built, governed, operated, and evolved ”.

The OECD has also raised concerns about competition, switching barriers, limited operability and high migration costs.

Those challenges are creating an opportunity for entrepreneurs who believe control itself could become a product.

British-based entrepreneur JB Benjamin is betting on precisely that.

Benjamin built Veldris to give organizations greater control over their AI infrastructure. Its platform, Midgard , is designed to give organisations more control over how their AI systems operate, from their data and computing to security, access and governance.

It can run on hardware owned and controlled by the customer, even in an isolated environment with no connection to the outside world. That means organizations can keep sensitive AI workloads under their own control rather than relying on an external provider.

“Convenience should never come at the risk of lack of sovereignty or privacy. We have become lulled into a false sense of security, trusting corporations and governments over ourselves and our own judgement. Unless you control all layers of your stack, there is no true sovereignty,” he says.

The underlying idea is that organizations should be able to see where their data goes, understand what software is running, control who or what can access it and move or isolate systems when circumstances require it.

Benjamin believes AI control will become a commercial product in its own right. His background in cybersecurity, digital identity and privacy has led him towards a problem that could become more important as AI moves deeper into business operations.

As systems evolve and become more autonomous, the infrastructure holding them together becomes even more critical. If an organization cannot properly inspect that infrastructure, its ability to govern the AI sitting on top of it is necessarily limited.

Benjamin is not alone in seeing the opportunity.

French company ChapsVision has built a substantial business around trusted data intelligence and AI, with around 1,000 employees and more than 2,000 large account customers. It operates across sensitive sectors including defence, intelligence and security as well as finance, healthcare and industry.

Its recent moves illustrate how sovereignty is becoming a commercial proposition rather than simply a government objective.

The French government’s Directorate General for Internal Security recently selected ChapsVision to provide a European sovereign platform for large scale processing of heterogeneous data.

The first phase of AI infrastructure was about obtaining enough computing power to train increasingly large models. The next phase may be about something more practical for enterprises: knowing exactly what is happening inside the systems they are paying to run.