What is sovereign AI?
A plain explanation of sovereign AI, data residency and private AI, written for organisations that hold sensitive data.
Last updated 15 October 2026
The short answer
Sovereign AI means you control the whole AI system: where the data is stored, where the model runs, who can access it and who can change it. Data residency, keeping data in the UK, is only one part of that. A service can store data in a UK region and still be run, changed and accessed by another company.
Residency and sovereignty are different
Data residency is about location: which country, and which data centre, your data sits in. Sovereignty is about control. Who can read the data? Who can change the system that processes it? Whose law can compel access to it? Who can switch it off?
The two are easy to confuse because suppliers often sell residency as if it were sovereignty. A UK region is a good start. But if the model, the software and the people with administrative access all belong to another company, you have residency without much control.
Three ways to keep control
Each step gives more control, and asks more of you.
A managed AI service in a UK region
The provider runs everything, and you get residency and contractual protections. You have little control over the model, its changes or who at the provider can reach it.
A managed model in your own cloud subscription
For example, an Azure OpenAI model deployed in your subscription as a regional deployment in a UK data centre, reached only over private networking. Prompts are processed in that region and your data stays in your tenancy; the network, identity and logging are yours. The provider still runs the model, under your agreement with them.
A model on compute you control
An open weight model runs on your own servers or on GPU machines in your own cloud tenancy. Documents, prompts and answers are processed only on infrastructure you run. The most control, and the most responsibility.
How much sovereignty do you need?
Match the control to the data, not to the headline. Internal guidance that is already public can sit happily on a managed service. Client confidential material, privileged advice, health data and research IP usually justify one of the last two options. Most organisations end up with a mix: a managed service for low risk work, and a private AI platform for the documents that matter.
Questions to ask any AI supplier
- In which countries are prompts, documents and outputs processed and stored?
- Are they used to train or improve any model, and can that be switched off by contract?
- How long is anything kept, including for safety or abuse monitoring?
- Which subprocessors are involved?
- Who at the supplier can access our data, under what process, and is it logged?
- Can we reach the service only over private networking?
- How are changes to the model or the service announced, and can we delay them?
- What evidence can our security team review?
- How do we leave, and what happens to our data when we do?
If a supplier cannot answer these clearly, that is your answer. Our AI Solution Architect service includes this kind of independent supplier oversight.
Start with an assessment.
An assessment shows which of your data needs which level of control.
Thank you. Your request is on its way. We will reply from info@tlgsolutions.co.uk to arrange a first discussion.
That did not send. Please email info@tlgsolutions.co.uk instead, and we will reply from there.