Own ground
Where your AI runs shapes how much control you have. We build in the Netherlands and show where we depend on others.
Choose for yourself, and be able to switch
Own ground starts with a practical question: can you carry on if a supplier changes its prices, terms or service?
You need to know who sees your data, where the work runs and what switching would involve. A server in the Netherlands answers only one of those questions.
Our machines, our responsibility
ITARR has its own machines in Breda running open language models. This lets us choose models ourselves and run work locally.
It also makes us responsible for maintenance, capacity and outages. Our own hardware has limits. A local model does not match an external service on every task.
The rest of the chain matters too
For other work, we also use AI services outside Europe. Our email runs through Microsoft 365. Some of our code is still on GitHub.
A model running here may still come from China or the United States. Chip design and the surrounding software are part of the dependency too.
That is why we publish a dependency map. It describes the current situation and possible ways out. A planned switch is not a completed move.
Start with the work
For each assignment we consider the quality needed, the sensitivity of the data, the costs and the ability to switch. That calls for a specific choice of services.
Want to know how this website works? We apply the same questions to it.