Behind the scenes at the SZ Summit with Noxtua & IONOS

At the SZ Economic Summit 2025 in Berlin, our CEO & Co-Founder Dr. Leif-Nissen Lundbæk discussed AI in Europe with Dr. Andreas Nauerz, Executive Board Member & CPO of IONOS, moderated by Jannis Brühl, Team Leader Money & Tech at Süddeutsche Zeitung. Our Chief Communications Officer Dr. Clara Herdeanu met with the panelists for a behind-the-scenes conversation.  

Clara Herdeanu: Why is everyone suddenly talking about digital sovereignty, and what does it mean for AI? 

Leif-Nissen Lundbæk: 

Sovereignty marketing seems to have become the new greenwashing – but just like with greenwashing, it is important to ask what sovereignty actually means.  In the context of AI, digital sovereignty essentially means three things: control over data, models, and infrastructure. If we hand over this control to non-European providers, we jeopardize the cornerstones of our democracy – from data protection to professional secrecy.  

This is an existential issue, especially in the legal sector. Professional secrecy is only a secret if we can prove that it remains one. With US infrastructure, this is simply not legally possible. Since the statement by Microsoft's General Counsel in the summer, we know that even he could not guarantee under oath that European data would not flow to the US. That is a fact, not political spin.   

Andreas Nauerz:  

Digital sovereignty primarily means the targeted reduction of dependencies on individual providers. The central element is interchangeability: organizations should be able to switch providers flexibly and securely at any time. At its core, it is about having full control over one's own digital infrastructure, one's own data, and the technological foundations on which our economy runs. 

Sovereignty is multidimensional.  

It is about legal sovereignty, i.e., the question of whether data is subject to the jurisdiction of foreign legal systems. Providers should be subject to German or European law. Otherwise, they remain vulnerable to geopolitical influence or laws such as the US CLOUD Act. If authorities abroad can access data in a non-transparent manner, this is not sovereign.  

It is also about technical sovereignty, i.e., the ability to further develop and operate critical hardware and software stacks ourselves – based on open standards and interfaces, rather than proprietary platforms that are effectively controlled by third countries.   

Ultimately, it's about operational sovereignty: the entire chain – planning, deployment, operation, monitoring, incident response – must remain entirely in European hands. If operations or customer data end up in third countries, that's not sovereignty.  

All three areas are crucial. Focusing solely on the geography of the data center is far too narrow a view. The decisive factor is whether control over the data centers also lies in Europe. 

Why does Europe need its own AI systems and not just an optimized ChatGPT?  

Leif Lundbæk:  

Because generic models neither understand legal domain logic nor meet regulatory requirements. ChatGPT is trained with data from the open internet—everything is included. Legal data is only a very small part of this, and if any is included at all, it is mainly legal data from the US context. This naturally also affects the results. One could even speak of the danger of an “Americanization” of the law. Another crucial point is that ChatGPT is simply not legally compliant because the data flows out. This ties in with the point about European control. We are therefore building a domain-specific AI system that is highly specialized in the legal context, which understands legal arguments and at the same time ensures maximum confidentiality. 

I think this is a democratic necessity and not an optional luxury.  

Andreas Nauerz: 

Furthermore, Europe should not copy everything that is developed in the US. Europe can do more than just be a copy cat! We must shape AI, not just use it!  

Europe's strength lies in the domain knowledge it has built up over decades, i.e., in its deep and diversified industrial, engineering, and process knowledge in fields such as automotive engineering, mechanical engineering, chemistry, energy, and medical technology. This knowledge is highly specialized. General language models cannot replicate this. While many US models are predominantly trained on open text and image data, we in Europe – especially Germany – can use huge amounts of high-quality industrial data (sensor data from production facilities, quality data, operating data, maintenance data, process chains). This is precisely the data on which industrial AI unfolds its true added value.  

Because AI value creation in Europe means using AI to produce products faster, more efficiently, and ultimately more cost-effectively, and using AI to make the products themselves more intelligent and thus more competitive. The economic leverage lies not in the next chatbot, but in optimizing a factory that has been running for 20 years. That's why we need compact, domain-specific, and energy-efficient models that can run on smaller “edge devices” and industrial machines and can therefore be integrated into machines, robots, vehicles, and production lines.  

The new game is called technology convergence, because the real disruption is not caused by isolated AI, but by the interaction of AI and autonomous systems such as robotics, ultimately resulting in robots that can perform complex cognitive tasks.  

The most interesting AI applications arise where models are deeply embedded in products, machines, or even the legal system. And that requires European domain knowledge. And, of course, all of this applies not only to industrial topics, but also to other fields, as Leif explained using the example of “law.” 

IONOS and Noxtua work closely together. In the fall, for example, we jointly launched Germany's first sovereign Legal AI Factory. What makes the partnership so European?  

Andreas Nauerz:  

We complement each other perfectly.  

Noxtua brings specialization and AI expertise, while we provide the European cloud infrastructure, hosted and operated entirely in Germany. The legal environment in particular shows how important this combination is: if the state or the judiciary are to use AI, the infrastructure must be technically, legally, and operationally unassailable. 

Leif Lundbæk:  

Partnerships are essentially the European way. We connect European legal publishers, specialized AI teams, and sovereign cloud infrastructure.  Without partners like IONOS, a purely European cloud company, we wouldn't be able to operate systems like Beck-Noxtua or MANZ-Noxtua in a legally compliant manner.  

And through joint GPU procurement, we even achieve a price advantage over US hyperscalers – so sovereign doesn't mean “more expensive,” quite the contrary.  

Often, it's also about fears: robot judges, AI decisions about life and death. How real are these fears?  

Leif Lundbæk:  

They are one thing above all else: misunderstandings. None of us want AI to decide on life and death, and that won't happen in Europe. 

But we have another, very real problem: when courts are so overburdened by a shortage of skilled workers and a huge workload that they are no longer even capable of making decisions, this poses a much greater threat to the rule of law. The state must be efficient – including, and especially, in the judiciary. In this sense, successful digitalization is also important for strengthening trust in the rule of law and in democracy. This also ties in with another point: namely, why digital sovereignty is so crucial in the legal system. Alongside defense and medicine, the legal system is one of the most sensitive areas of a state. An independent legal system is also a guarantor of the rule of law and thus, indirectly, of democracy. Digital sovereignty in the legal system is therefore indispensable.   

AI can automate routine tasks in the legal system, provide assistance, and make workflows more efficient, but nothing more and nothing less. Humans remain centrally “in the loop,” especially in difficult cases.  

Andreas Nauerz:  

In our view, the most dangerous scenario is a world in which Europe has lost its digital sovereignty.  

If we do not know what training data an AI model is based on, what criteria it is curated according to, and who ultimately controls it, then others effectively determine what is considered true or false, right or wrong. In such a constellation, we are talking about the perfect propaganda tool. This is because AI can generate, amplify, and scale enormous amounts of content in a very short time.   

This is highly problematic, especially against the backdrop of the current geopolitical situation and developments in individual countries. As political influence on educational content, media, and public discourse increases, control over AI systems becomes a direct question of power. AI is then no longer just a technological tool, but a strategic instrument for controlling perceptions, opinions, and social debates. 

In a few years, people will no longer obtain most of their information via traditional search engines, but directly via AI-based systems. It will be completely normal to no longer use Google, but to consult ChatGPT, Gemini & Co. directly. However, this shifts the actual control over what information we receive, what narratives emerge, and what is considered true or false to those who train, operate, and control these models.  

This is precisely why controllable European AI solutions are not a nice-to-have, but a must-have. 

And where is AI heading overall? Will it change everything, or will it be slower than expected?  

Leif Lundbæk:  

Both.  

Social change will take longer than some people think, but it will be massive.  

AI influences our language and thus our thinking. That's why we can't say, “We've lost the race, so now it doesn't matter.”  

Europe must develop its own models, including large ones. The impact on democracy and the rule of law is too great to become dependent on it.  

Andreas Nauerz: 

Its spread will be exponential, but it will raise many ethical and social questions.  

And we are only just at the beginning.  

What many people currently underestimate is that we are not only experiencing a technological leap, but also a democratization of the entire technological landscape. AI will increasingly break down the traditional divide between “experts” and “users.” In a few years, we will be able to control the most complex systems without any technical expertise, using only intuitive voice and dialogue interfaces – whether simple devices such as ovens or washing machines, highly complex IT systems, or autonomous robots.  

This will be accompanied by the next major stage of development, the technological convergence already mentioned. AI, robotics, automation, cloud, and edge are converging into a unified sphere of activity. We will tell robots what to do in natural language – and they will understand and implement it. AI will thus evolve from an analysis tool to an operational control center for physical and digital systems. 

This will change entire value chains – faster, more fundamentally, and more broadly than many people expect today. Not gradually, but structurally.  

Clara Herdeanu: That's a very apt conclusion. Dear Andreas, dear Leif, thank you very much for the interview and your plea for true European digital sovereignty!