In 2025, McKinsey surveyed 300 executives and officials to find that 71% see sovereign AI as no less than an “existential concern” or “strategic imperative”. Similarly, Deloitte has found that 83% of organizations surveyed in its 2026 “State of AI in the Enterprise” report see sovereign AI as moderately important to their strategy. There is plenty of other evidence that sovereign AI is gaining traction globally, and it goes beyond business goals or data privacy matters. Geopolitics say hi.

So what is sovereign AI and how does one build a stack that works for their IT, keeping compliance in check while delivering practical confidence and, what it all started with, the new level of efficiency and productivity? Let’s find out.

Data sovereignty and AI: Can you have both?

Governments are already backing sovereign AI with a lot of money. As Fortune reports, in 2026 more than $100 billion goes into sovereign AI compute. Gartner, too, finds global sovereign AI one of the top trends in data and analytics in 2026, and connects it with “a nation-state’s ambition to develop and use AI, without foreign assistance, for sovereign objectives, including global AI leadership and economic competition.” According to the Sovereign AI Index by CNAS, countries in the Middle East and East Asia account for more than 80 percent of all tracked and publicly disclosed sovereign AI investment worldwide.

There is an opinion that the borderless cloud era is over. Driven by data residency rules and supply chain risk, there’s a shift toward domestic-first model selection, hosting, and procurement.

Nonetheless, there’s an interesting paradox found by CNAS and refllected in their Sovereign AI Index: most sovereign AI projects depend heavily on foreign (and mostly U.S.) technology.

To decide whether truly sovereign AI is possible today, we need to understand what it actually means in practice.

What is sovereign AI

McKinsey offers four dimensions to define the spectrum of sovereignty:

• Territorial: where data and compute physically reside
• Operational: who manages and secures data and compute
• Technological: who owns the underlying stack and intellectual property
• Legal: which jurisdiction governs access and compliance

It argues that sovereign AI is about how the whole AI stack is controlled: how it’s created, trained, and deployed, spanning infrastructure, models, and applications. Such an approach reminds us of our very own Ethical AI Rating that considers open source licensing, availability of training data, and availability of self-hosting options to be three key characteristics of an Ethical AI model.

As Nextcloud founder Frank Karlitschek pointed out in his interview for the 10th anniversary of Nextcloud, digital sovereignty is when:

• You know where your data is
• You can see how the software works
• You can change and adapt it

So it all comes down to control over your tools that you can verify and act on, not just control in theory. Similarly, sovereign AI is when you decide where your data is located, you know how models that process it work (including how they are created and trained), and you choose what AI features and components to use.

Now, do you think AI can be part of your digital sovereignty strategy?

Key success factors for Sovereign AI are built into Nextcloud Assistant by design

Nexcloud offers the first AI assistant that doesn’t prey on your data. We designed it to meet every criterion that matters for a sovereign AI infrastructure:

• You control the location of your AI and data by choosing where to host your AI: on your servers or with a provider you can trust in the jurisdiction you prefer.
• Information about your AI models and tools is available to you. Open source AI tools and open, ethical AI models guarantee the transparency you need to build a sovereign AI stack. Here’s where our Ethical AI Rating is especially helpful.
• Open standards and interoperability. You choose what AI features to use and what models they are powered by, without being locked into the ecosystem of a single Big Tech vendor. 

Clearly, sovereignty means choice. One more important thing is that AI in Nextcloud Hub is fully optional and disabled by default. If you prefer sovereignty by staying completely AI-free, Nextcloud respects that.

Sovereign AI agents in Nextcloud Hub

Nextcloud Assistant offers agentic features to enable automations in routine tasks across many Nextcloud Hub apps. AI agent features let it perform actions with your apps and data and take part in automating complex, nuanced workflows. 

For example, you can ask it to create a calendar event, send a chat message in Nextcloud Talk, or check the weather for you. Or do something more complex, like look up and chat with a customer, find out what they are looking for, and generate a template for a slide deck for a tailored presentation. All that without risking any leaks of your sensitive conversations and documents to a third party.

EU AI Act: How do I stay compliant?

Sovereign AI is not only about how you can make your AI tools more secure and private. Taking into account the global state-level trends we talked about earlier, it is also very likely becoming imperative rather soon.

Take, for example, the EU AI Act. It requires detailed information on where and how features are integrated. Nextcloud is fully prepared for this, since the information is available in our online documentation portal.

In our latest release Nextcloud Hub 26 Spring, we also added more information on the AI integration that is used in the background. For example, when using the AI Assistant, every task will show which AI is in use: whether it is a public AI model or the service using the IONOS AI app, for example.