A conversation with Dr. Leif-Nissen Lundbæk, CEO and Co-Founder of Noxtua, about the beginnings of Noxtua, the path to digital sovereignty, and the future of the European legal system.

“We must not allow others to decide what is true or false in our legal system. This is not a technical question, but a democratic one.”  

Interview: Dr. Clara Herdeanu (Chief Communications Officer)

Leif, the idea for Noxtua was born back in 2017. What was the catalyst at the time?  

In 2017, I was researching AI and data privacy for my PhD thesis at Oxford University and Imperial College London with my supervisor, Professor Michael Huth. So initially, it wasn't primarily about Legal AI, but more about the basics: How do we build AI systems that serve people without having to give up control over our data? Or to put it another way: How can we develop powerful AI based on European values such as data protection and transparency? Privacy by design was still quite a niche topic back then; today, everyone is talking about it, but in 2017, things were different.  

The initial spark for Noxtua as a company (then known as Xain or Xayn) came from Oxford, but we ended up founding it in Berlin. We now have our headquarters there and in Paris, with additional locations in Zagreb and Munich. This actually also demonstrates how we think: European, right from the start. Because we are convinced that certain principles, such as privacy, transparency, and independence, only work if you take them seriously from a technological perspective.  

And then the focus shifted to the legal sector?  

Noxtua, Europe's sovereign Legal AI, is now officially two years old, and we are very proud of everything we have achieved during this time. We officially launched the first version of Noxtua at the end of February 2024, but we had already been working on it behind the scenes for several months. A little over a year later, we were able to announce our close partnership with C.H.BECK, Germany's leading legal publisher, in what was then Europe's largest financing round for Legal AI.   

This focus on the legal sector was a logical consequence and further development for us and me, following years of intensive work on AI, data protection, and security. It is also worth emphasizing here that we have not only been working on these topics since the ChatGPT boom, but have been developing AI in great technical depth for years.  

When you think about the areas in which privacy and digital sovereignty are truly existential, three points quickly come to mind: defense, medicine, and the legal system. An independent legal system is one of the pillars of the rule of law and thus also of democracy. If we relinquish control here—over data, over model systems, over infrastructure—we ultimately undermine the the rule of law itself.  

At the same time, however, the legal field also offers tremendous potential for AI. It is a highly structured, rule-based system based on clearly defined texts. But that is precisely why you cannot simply use any generic model here. Legal precision is non-negotiable. AI must understand how legal thinking works, and that requires trustworthy legal data, not Wikipedia or Reddit. 

You talk about “digital sovereignty.” What exactly does that mean?  

The wording “digital sovereignty” is now used inflationarily, almost like greenwashing, but for tech. That's why it's important to be specific about what this actually means. For us, digital sovereignty means three things: We must have control over the data, over the AI systems, and over the infrastructure. All three. If even one of these areas is outside Europe, it doesn't work.  

For us, this means that everything runs entirely on European, sovereign infrastructure. For example IONOS, Open Telekom Cloud, and no US hyperscalers. To this end, we have data centers in Magdeburg, Berlin, and Munich in partnership with these European companies. With IONOS, we have also set up the first sovereign Legal AI Factory in Germany in Munich.  

And I also want to stress this point: “European infrastructure” does not just mean that the servers are geographically located in Europe. They must also be operated by European companies and subject to European law. This may seem like a small difference, but it is actually a huge one. In other words, not everything that calls itself sovereign is actually sovereign.   

So pure geographical location is not the main criterion. Why is it not enough for a US provider to have its servers in Frankfurt, for example?  

Geography is not synonymous with control. And that is no longer a secret. In the last summer, Microsoft's GC testified before the French Senate—under oath, mind you—that he could not guarantee that European customer data would not flow to the US. That pretty much ended the discussion.   

For lawyers, this is a massive problem. Professional secrecy is only a secret if you can ensure that it remains one. With US infrastructure, this is legally impossible. Anyone who ignores this cannot later excuse themselves by saying “I didn't know.”  

That's why we at Beck-Noxtua, MANZ-Noxtua, Swiss-Noxtua, and all other variants of our Legal AI Workspace comply with all data protection and professional law requirements of the respective country.   

That sounds complicated. Why not just use the existing tools?  

Because the existing tools are the problem, not the solution. And we have decided not to be part of the problem, but part of the solution.   

Yes, it is more complicated to set up your own infrastructure at the beginning. But what you get in return is security, control, and ultimately better performance. We see it every day: the AI is faster because we're not competing with a thousand others for computing time. It's more accurate because it's based on legal data. And it's secure because the data never leaves the country.  

Those who rely on hyperscalers may save themselves work in the short term, but they give away control completely. And interestingly, joint GPU procurement with partners such as IONOS is even cheaper than with the big US providers. Sovereign does not automatically mean expensive. 

Why is a generic AI such as ChatGPT not sufficient for legal work?  

Because ChatGPT is a generic model, trained with potentially everything that is online on the internet. That may be okay for small talk or brainstorming, but not for the legal domain. Legal content makes up only a tiny fraction of the training data, and if anything, it is mainly US law. This naturally colors the results. One could almost speak of a creeping Americanization of the law. 

But the bigger problem is another: ChatGPT is not legally compliant. The data flows out. Period. This makes it simply unusable for professionals bound by professional secrecy, no matter how good the interface looks. 

We need domain-specific model systems that understand how legal reasoning works.   

How can you ensure that AI can “think in a legal way”?  

By working with the right data. Legal AI is only as good as the database behind it.  

That's why we work with leading legal publishers from all over Europe. In Germany, for example, with C.H.BECK, Germany's most important legal publisher. Beck-Noxtua is based exclusively on beck-online, which contains more than 60 million curated documents that comprehensively cover the German legal system. No other AI providers have access to this. This is a core part of our competitive advantage. 

Our AI therefore works with legal texts, commentaries, judgments, and legislative texts, rather than just any websites. This enables it to understand why an argument is conclusive, not just that it appears somewhere. And it can transparently show which sources it refers to.  

In addition, there are strategic expansions; for example, in Germany, we are integrating content on public service law with Erich Schmidt Verlag. And with Walhalla, we are expanding the area of administration. Each partnership makes the system better and broader. 

Is Noxtua also expanding throughout Europe?  

Absolutely. The European idea runs through everything we do. That's why we recently became the first provider to launch the European license. This allows legal professionals to access multiple publishers and jurisdictions across the continent via a single interface. We have already launched this in the beta phase for the DACH region and are now gradually adding further publishers and jurisdictions.    

This is precisely our European approach: we want to enter into partnerships, respect national characteristics, but at the same time create a common ecosystem. 

What can we expect in 2026?  

In my view, 2026 will be the year of the justice system. The courts are under enormous pressure—staff shortages, exploding case numbers, and on top of that, they are being inundated with AI-generated briefs from law firms. They will have no choice but to rely on this technology themselves. And that's where we want to offer the right solution.  

At the same time, we are continuing to expand our European network. Beck-Noxtua, MANZ-Noxtua, Swiss-Noxtua – these are the first building blocks. But it's about more than individual products. We are building a European ecosystem for Legal AI. A system that proves that digital sovereignty is not just a vision.   

What is the long-term vision for Noxtua?  

We are building Europe's sovereign Legal AI. An AI system that shows that powerful AI can be developed based on European values, data protection, transparency, and the rule of law. Because, in our view, we don't simply have to replicate everything that comes from the US.   

Our goal is to connect all jurisdictions in Europe with specialized legal information—and we want to achieve this by the end of the year. To this end, we have introduced the Europe license, which allows lawyers to access multiple publishers and jurisdictions via a single platform. We started in the DACH region and are adding more and more legal publishers and jurisdictions from across Europe in the coming weeks and months. So there is still a lot happening at Noxtua this year.  

We are therefore deliberately positioning ourselves as a European company. In our view, Noxtua is not only an AI solution, but also a European project. We want to build a European ecosystem in which national legal systems are respected, but at the same time the advantages of a common, sovereign infrastructure can be exploited. With Noxtua as a beacon of sovereignty, we also want to show that Europe can create great AI. A Europe that shapes its own digital future, especially in the legal field, where independence is so fundamental.  

After all, AI influences our language and thus our thinking. That is why we cannot say, “We lost the race, so now it doesn't matter.” Europe must build its own AI systems, including large ones. The impact on democracy and the rule of law is just too great to allow ourselves to become dependent. 

That's a great way to end this interview. Thank you very much, Leif!