Rise of AI Conference 2025 · Berlin · Applied AI Stage

Unblocking Bureaucracy

Georg Schlenkhoff, Oliver Stuke · Georg Schlenkhoff: Mistral AI (go-to-market lead, Germany); Oliver Stuke: Capgemini Invent

At Rise of AI 2025 in Berlin, Georg Schlenkhoff (Mistral AI) and Oliver Stuke (Capgemini Invent) argue that Europe, despite a widespread narrative of having fallen behind in AI, is building real momentum toward sovereign generative AI, and show how that stack can be applied to unblock public-sector bureaucracy. Schlenkhoff frames sovereignty as owning infrastructure, technology and data, citing figures that only 3 percent of European R&D spending goes to technology and software and that Europe consumes 20 percent of global chips while producing only 9 percent. Stuke points to recent political momentum, including a German conference of minister presidents' sovereignty declaration, a Dutch parliamentary declaration and the EU's 200-billion-euro continent AI action plan, and presents a layered sovereign stack built with partners including StackIT, Aleph Alpha, Mistral, Fraunhofer, Demos and Atlas Systems. They walk through a concrete application in Germany's Ministry of Interior, where an AI system checks planning applications for wind turbines against legal requirements and drafts notices to applicants, while stressing that German administrative law bars fully autonomous decisions and requires a human in the loop. Audience questions cover adversarial attacks, change management and ethical data handling.

Key takeaways

  • Schlenkhoff cites figures that Europe captures only 3 percent of global R&D spending in tech and software, and produces only 9 percent of the chips it consumes.
  • The EU's continent AI action plan is funded with 200 billion euros, covering chip gigafactories, talent training, and public- and private-sector adoption.
  • Their sovereign AI stack combines StackIT (compute), Aleph Alpha (explainability), Mistral (frontier models), Fraunhofer (compliance research) and Demos (citizen participation).
  • German administrative law (article 34b) forbids fully autonomous administrative decisions, so their planning-permit AI system keeps a human clerk in the loop.
  • Schlenkhoff projects a shortage of 1 million skilled public-sector workers by 2030 as a key driver for applying generative AI to reduce mundane workload.

Detailed notes

Is Europe really behind in AI?

Schlenkhoff and Stuke open by challenging the widespread thesis that Europe has lost its track in AI. Stuke argues the digital realm has only recently become a real field for European geopolitics, after years of governments focusing on regulation rather than treating technology like other strategic resources such as steel or coal. Schlenkhoff defines sovereignty as owning infrastructure, technology and data, which matters because governments handle sensitive citizen and corporate data, and cites figures that only 3 percent of European R&D spending in technology and internet software stays in Europe, while Europe consumes 20 percent of global chips but produces only 9 percent.

Political momentum toward a sovereign stack

Stuke lists recent political signals: the German conference of minister presidents adopted a declaration calling for sovereign AI and less dependence on US technology providers, the Dutch parliament passed a comparable declaration asking for a sovereign stack, and the EU declared a continent-wide AI action plan funded with 200 billion euros covering gigafactories with 100,000 chips, talent training and public- and private-sector adoption. Schlenkhoff adds that the new German coalition agreement similarly calls for European, trustworthy digital providers and reducing digital dependencies, summarized as building a "German stack" that must remain interoperable across the EU.

Building a sovereign AI stack from real components

Drawing on the Euro Stack initiative, Stuke describes a layered stack running from raw materials and chip production (citing ASML) up through AI operating systems and models to the application layer. For their own solution, the pair assembled StackIT for European GPU compute, Aleph Alpha for explainability, Mistral for frontier models, Fraunhofer for trusted-AI and compliance research (operationalized by Capgemini Invent), Demos for AI-augmented citizen participation, and Atlas Systems for end-to-end confidentiality and security as an outer layer on the StackIT infrastructure. They present this combination as a first working example of a sovereign AI stack rather than a theoretical concept.

Why generative AI, not replacing workers

Schlenkhoff identifies three drivers for applying generative AI in government: a projected shortage of 1 million skilled public-sector workers by 2030 as staff retire while workload grows; approval procedures that studies show run six months longer than legally required; and a 500-billion-euro government funding program whose bottleneck, he argues, is not money but slow planning processes. He frames generative AI as freeing staff from repetitive tasks to focus on stakeholder relationships, not as a replacement for them, citing legal, data-bias and interpretation challenges including the EU AI Act's high-risk classification rules and GDPR, plus the need to keep humans in the loop and decisions transparent and explainable.

Case study: automated planning-permit checks

The pair demonstrate a real deployment with Germany's Ministry of Interior, where a company applying to build a wind turbine submits an application that the AI system checks against the relevant environmental law (BImSchG paragraph 10), extracting a legal checklist and mapping it against the submitted document and metadata, including blacking out personal data. The system produces a to-do list for both government workers and applicants and can draft a notice, for example confirming a site plan and noise forecast are acceptable while flagging that an EIA summary and public-insurance proof are still missing, with 21 days to submit them.

Q&A highlights

Asked about adversarial attacks using fabricated but fully compliant applications, Stuke says the goal is to mirror real-world verification processes into the digital system, and notes German administrative law (article 34b) explicitly forbids full automation of these decisions, so a human clerk remains in the loop. On citizen-participation processes being flooded with AI-generated submissions, Schlenkhoff points to Demos's tooling for detecting and clustering AI-generated input. On change management in inflexible public organizations, Stuke cites an eight-year German Autobahn digitization contract where change management was built into the tender from the start, and both stress involving end users early. On ethics and data protection, Schlenkhoff argues sovereignty in deployment, running open-source models within secured, self-contained infrastructure that does not communicate externally, is how the approach keeps sensitive data from leaving secured premises.

Quotes

Sovereignty from a nation's perspective means that you are master of your own destiny.

Georg Schlenkhoff

Any technical solution cannot stand on their own feet.

Oliver Stuke

We embrace it and we say it's good because we really regulate the main parts.

Georg Schlenkhoff

Questions answered in this talk

Is Europe behind in the AI race?

According to Georg Schlenkhoff and Oliver Stuke at Rise of AI 2025, despite a common narrative that Europe has fallen behind, there is significant political and technical momentum toward sovereign European AI, including a 200-billion-euro EU continent action plan and national declarations in Germany and the Netherlands.

Can generative AI replace public-sector workers?

Schlenkhoff and Stuke state at this talk that generative AI is meant to remove repetitive, mundane tasks from public servants, not replace them; they note German administrative law (article 34b) explicitly forbids fully autonomous administrative decisions, so a human remains in the loop.

How can AI speed up government approval processes like building permits?

Stuke and Schlenkhoff describe a deployed system for Germany's Ministry of Interior that checks planning applications, such as for wind turbines, against legal requirements like BImSchG paragraph 10, generates a compliance checklist, and drafts notices to applicants about missing documents.

Mentioned

Sovereign AI, generative AI in the public sector, human in the loop, BImSchG planning permits, digital sovereignty, citizen participation AI

Keep exploring

More from Georg Schlenkhoff

More from Rise of AI Conference 2025

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 5,213 words

Show the full transcript

Yeah, good afternoon everyone. Um, thank you for being here on stage. We will not focus on cars today but more on the public sector. We will think of how can we apply such a transformative technology like generative AI in the public sector. And we want to focus on one key aspect in how can governments keep their interests. And this brings us all to the aspect of sovereign generative AI. But if we talk about sovereign AI or AI made in Europe, there often comes the thesis maybe Europe has lost its track in such an important technology. Oliver, what's your perspective on this question? Yeah. Well, one might think so, but let's have a look at the conference.

Well, the model of it is good news from Europe and I think rightfully so. There's a lot of stuff going on, most of it actually into the right direction. And let's use the next 30 29 minutes to explore why we think that's the case. Um, and why the good news might be even better uh than we might have thought in the uh uh beforehand. But before diving into this topic, uh let us spend a few more minutes of seconds uh introducing ourselves and why we're talking about this topic. Well, now you've just heard who we are. I have to press one of these buttons. Actually, probably some of them. Let's see. Not that one. Another one.

Is this actually on? Let's see. Well, um, while we're figuring that out, um, as we've heard, um, I'm Oliver from Cup Gemini invent um, and as the leading consultancy with a European heritage, we don't only believe in Europe. We are committed to purpose and we are actually driving forward. Oh, great. Uh, 15 years of experience in the digital sector. That helps actually. Okay. So we are actually driving Europe's digital sovereignty forward. Uh and we'd like to explain you in in the ensuing minutes how we actually doing this and as so often ambitious targets are not achieved best by doing it on your own but with the help of good friends and partners which is why I'm delighted to share the stage with Gayok today.

Yeah thank you Oliver. Yeah I'm Gok I'm leading the go to market for Mistral in Germany. So for uh those who don't know Mistral, so good. Mistral is one of the leading um state-of-the-art model providers located in the heart of Europe. So we are based headquartered in Paris and uh yeah we develop large language models but also solutions for geni and a platform for geni. We are committed for open source um to to open source because we feel open source is one key component for public sector innovation because at the end you want AI engineers to understand what the models do, understand the architecture. You also want developers to adjust the models. This could mean fine-tuning or train them new languages like the German public law or maybe bump dodge.

And finally, you also want to enable agencies to distribute solutions on their terms if and this is only possible if you base the solution on open source. So that's why we are committed to open source. But let's dive into uh action what what it let's get back to our topic. What does this mean for uh the digital sovereignity of Europe? So why do we feel uh this is currently a topic? Let me just briefly define what what means sovereignity. So sovereignity from a nation's perspective mean that you are master of your own destiny. Precisely you own the infrastructure, you own the technology but you also own the data and and this is because we we handle with sensitive data in and of citizen and and also firms on government level.

So this is a very critical point and this is even more critical if we consider two numbers. So from a recent study of of battlesman it was said that only um 3% of the R&D spendings for uh technology and also internet software is spent in Europe. So that's one one problem and on the other side we see for example like 20% of the chips are consumed here in Europe while I only 9% of the chips are produced here in Europe. So Oliver we wonder how does do these numbers like change the mindset of the policy makers. Yeah. And and that's likely one of the big questions today in in our times isn't it? Well, so what what we're currently seeing is at least that's our impression that the digital realm hasn't been a real playing field for geopolitics at least from a European perspective uh in in the last years and decades and many well um commentators were wondering why that might be the case why the um interplay or the relationship between European governments

and the technology sector was so much focused on regulation instead of using the digital space as a field equally Y to other resources like cars, steel, coal and so forth for geopolitics. And right now it seems as if there's a major shift happening where European politicians adopt this new playing field. But well that sounds like a feeling. How does that materialize? Well, actually uh we look at the newspaper. That's a great way to start every every research. And we saw great momentum in every level of of governments from from the federal level where the uh German conference of minister presidents uh uh adopted a um declaration asking for sovereign AI and a bigger independence of um US technology providers to the um national level where the Dutch Parliament just a few weeks ago adopted quite a comparable um declaration asking their government to deploy a sovereign stack um for for the country to use and then also on the so level uh so national level where the European Union itself um has declared their um

EU uh continent action plan many of you likely all of you have heard about that and what does that entail it's uh it covers everything that we might need in Europe for a stronger um sovereignity around AI from building and constructing gigafactories with 100,000 chips to um training and retaining talent to actual adoption in the private uh in the private and public sector and all of that funded with 200 billion euros. Um but just uh well we we are in Berlin actually and just one week after the inauguration of a new government. So G what's their take on them? Yeah so their take is quite similar. So if we read through the German coalition agreement um so there's a a strong focus on European trustworthy digital providers also they ask um yeah to protect and expand our digital infrastructure here in Germany at the same time reduce the digital dependencies by establishing technologies like geni here in our country and then this all concludes uh to one let's say like buzz word it's it's comes

down to a German stack. So we wonder a bit h how can we demystify such a sovereign stack? What what does it mean? And even since it's also asked to make this st interoperable with Europe because at the end we are one European Union, how can we still do politics across the borders as well? Oliver, how do we see this demystified also a bit in the Euro ST? Yeah, that's the great question now. So um politics have set the guard rails or the guidelines where we want to go in the next four years. Um but build well filling that with life is now a task for the whole society and us as the tech and AI ecosystem isn't it?

We cannot just sit back and wait for governments to solve this conundrum on our behalf. And whenever we look at a complex issue what do we do? We we layer it down and solve layer for layer. That's actually what we're seeing with uh with a sovereign stack as well. Uh here we took the um example from a great initiative Euro stack of which we are a great uh proud member. And uh what they did was um building the whole stack from the bottom up with concrete solutions. And and if you look at the whole stack, it starts with sovereignity around raw materials which you need need for the um uh chip production. It starts then with um chip factories like ASML in Netherlands.

But it goes well beyond the the physical layer to the soft and logical layer which you see on the right hand side. um that includes AI operating systems, AI models, AI uh algorithms and then ends in the um in the say application layer where you use use cases, talent and so forth. But having a box full of single solutions doesn't actually help, does it? What you need is building connections between all of these elements and components to um then have solutions, modular ones surely, that you can apply to your specific um problems. Do you always need the full sovereign uh stake? Likely not. You do not need new raw materials for your next AI use case.

But some use cases require more of sovereignty, others less. And one use case or one one playing field where a where sovereignty is of increasing importance will now be um presented by Gok. Yeah. Yeah. So let's uh look at the application uh of of geni in the public sector because this of course is a vision that's a techn technology stack but we need to apply it and um in conversations uh with with public sector also sometimes with private sector we often get the notion or the question is oh will this board replace the humans so this is I think a real fear uh that we hear quite quite a lot and yeah we three drivers where we see it's geni is not a problem but it's really a solution uh for for what we are facing.

So if we look at 2030 actual studies say that 1 million skilled workers are lacking in the public sector because there's a lot of people going into pension and at the same time the amount of workload gets higher and higher. So this means more work, less people and you can um conclude that people get burnt out have have stress at work. So here GI is really helping to take away the mundane tasks the repetitive task but help the people to focus on what really um yeah they where they can use their brains. A second driver we see is that the approval procedures um also according to a study are six months longer than required by law.

So the question is is that an attractive market to invest in if that's the case. So here we also see that Jenny can help to really accelerate the approval and planning processes. This is one concrete example that we're also working at right now. And then um the third uh part is um if we look at the funding program of the government of 500 billion. I think the problem is currently not the money that that lacks for our for our country but how can we improve or accelerate the the planning processes because at the end we want prompt administrative decisions uh in the best case that yeah stakeholders that put in the application have a direct feedback.

Is my application complete? Uh where do I need to um maybe a bit leaz more with the stakeholders on the government side? with geni we can really help the governments focus on building relationships talking with their stakeholders with the uh requesters instead of having processes stuck right so but if we if we find that geni is a perfect solution to tackle these three main challenges it doesn't come without challenges in the deploying of that technology let us be precise what are like the key challenges we see in our daily lives in in the projects. Um so of course one main driver is compliant with law. We have the EU AI Act which some people say it's it's a major burden for our jurisdiction.

But on the other side we embrace it and we say it's good because we really regulate the the main parts. we regulate when is it um a high-risisk AI and when is it not an high-risisk AI and what do we need to apply in terms of further security mechanisms if it's a high-risisk AI same with GDPR how can we make sure that the personal data especially our citizen data do not go into the wrong hands then data bias is a big topic um if we train models with hate speech at some point the model might take a decision that goes against the values of what Europe makes or what Germany makes about. So we need to make sure that the the yet the training data is transparent and correct of course at the end and this is by the way also regulated by the EU AI Act.

And then we see fuzzy legal terms. So what does it mean? I mean if we have like the law with one sentence and we have on the other side an application it could mean that like three requirements come from one sentence. How can we make sure that the AI really understand this? and especially also makes this transparent and explainable. So that's also a problem we need to solve. And this brings us also to the point that the AI alone won't solve the problem and won't um accelerate the processes but only hand in hand with the human. So we need to put the human in the loop make him transparent what happened uh what was the decision the II met and suggested and how can we then put this into practice.

So the question Oliver is how do we solve in in our partnership such problems? Yeah and it's a great task um because as you see there are numerous challenges all all hard enough on their own but coming together um well they seem like like a huge conundrum. Um and while there are solutions available for each single challenge and as as we already said that in in the discussion of what a stack actually is um they they they cannot succeed if they stand individually. So what we did was we had a look at all of these challenges. We looked for the best fitting solutions with a catch. I'll come to that in a second. And we brought them together and we made them interoperable.

And uh let let's discuss those uh because you see all of these logos maybe you recognize most or all of them. Um but but well how did we do that actually? Um first we needed a European infrastructure to build on. Um that was stackit in this case with all the GPU compute power that we needed to to solve these tasks. Um then we have different single components like Alf Alfa who are masters in explanability as maybe all of you know and we have Mistral who are bestin-class regarding frontier models but what's the connection now? Well, imagine we apply the explanability features of alf alpha to the frontier models of misol. What we could achieve if we did that?

Yeah, we did that. Um, another aspect is we have uh foofer with uh leading research on trusted AI and AI compliance. But well, it's research focus. What we did was taking their results and their their great frameworks and applying them operationalizing. That was our task as as Cubge Gemini to um ensure that we tackle all of these challenges within the solution. Uh another aspect is um maybe not everybody's aware of that but there are someone who called demos who who are great in uh citizen participation with AI. Um and now we um augmented that and amplified that with these frontier models which became explainable and so forth in order to tackle that as well. Um and what are we still uh missing?

Well, Atlas systems as uh experts in end-to-end confidentiality and security were now applied as say an outer layer deployed on the on the um uh stackit IT um environment to to capsule all of this. Uh and as as as you discussed early on, what what's the catch? Well, all of these are sovereign in themselves or for themselves and all the oval solution has that say stamp as well. And we now have a a first stack um that that covers sovereignity, Europeanness, if that's a word. Um and helps us uh uh while solving the uh um the the the issue at hand with all of these challenges. and um can be seen as a first step of uh of a sovereign AI stack applied in the real world.

And we started obviously with with with the field where we had the biggest leverage up front with with the eye playing field. Now the other components can ensuingly be built on top of that and with that vibrant ecosystem where we also saw our role as Geminina to bring these players together. uh we can go ahead and solve the next uh uh issue for governments and well beyond that wherever sovereignty is key. I hope that makes sense. Yeah. So um what can we take out of that? Um so we still have a few minutes left. So maybe we can have a look at the example uh precisely of of of because we are building this not uh on paper but also in reality.

Uh for example we are engaged with the ministry of interior. So how how this looks like you know in in a real German government process where planning and approval processes is key to leverage investments. Um we we have here brought one of the um examples that where we apply also the STE. So imagine uh that there's a company that wants to build um a vint bin somewhere um on the German uh field and they need of course to hand in an application and make sure that according to the environmental regulation um the for example bimski that the application is complete. So what the model can do but only as said in combination of of a stack.

So where we make sure that the data are secured that we are that we can also um yeah black out personal data for example. Uh it will first look into the legal um documents in the law and find out okay what is the legal checklist if I apply bimski um abstract 10 to this application. So it will come up with a checklist and then extract on the other side the document that was submitted including the meta data that you can put on top um of of the of the document and map now the document on the one side with a legal checklist on the other side. So quite straightforward. Um it will come up with a to-do list that we of course need then to supply to the government workers but also uh to the requesters of the application to make this process very smooth and help to yeah come uh over lengthy periods of of checking uh such an application.

And finally uh the AI can even draft a notice for the applicant making sure okay the site plan we have uh checked this looks good the noise forecast is also fine but we are missing the EIA summary you have 21 days to hand this in so please move forward same with the public insurance proof so that's a concrete example um how the models can work in practice in in the government okay with that Being said, thanks for the example making everything tangible. Um, what's what can we take out of that on on a meter level? What what made Europe strong in the first place? Because that's the model of our conference here. It was setting aside differences and was cooperation under one mission, wasn't it?

And that does not only need to apply to government, it can also apply to us as the tech and AI ecosystem. And if we do that and you see we did that uh if we bring together these diverse players we can achieve so much more than the sum of just those individual pieces um for the sake of all of us. Um let's continue with that shouldn't we? The the well the conference model is good news from Europe. We hope we show they are even better. Um we have started building a applicable and already applied AI stack with full sovereignity uh tackling all of these uh issues and if if you want to let's cooperate on that together and uh yeah move forward.

Thank you. Um, any questions from the audience? Otherwise, I obviously have one. Hang on, wait for the microphone because we also have another hundreds of people in the audience online. Thank you. Um, I'm I'm Thomas Wicker. I'm normally doing cloud security. Uh, and here I have an example. uh if I do an adversary adversarial attack on this process for example drafting an input document which is completely fulfilling out the requirements but is completely fake what you detect it I mean the goal is um to mirror the real world so what happens in the real world we translate into the digital world so I would take back this question is how do the how is the process shaped today?

Does it detect such a um attack and then of course we need to implement the same mechanisms uh in the digital system. It's like a digital twin I would say right and maybe I can add to that if if you want me to. Well um obviously our main task is to build a an AI solution that can automize critical and time consuming tasks. that does not mean that we replace the administrative clerk. actually uh what's that article 34B of German administrative law even forbids a full automization of these um of of these processes which is why what what what go just uh told you um if right now an administrative clerk sees this application and he does several tasks um well we help him doing this quicker but it's not as if the the the application would be run through the system automatically autonomously without any interaction with a human.

So maybe that's the clearest point where your attempt would be well interesting but not well successful. Does it make sense? I don't think so. But let's talk. Okay. Well, we if if nobody else has question, we can talk about that in full length. Yeah. Because for example the the British unions applied uh this to uh to a working law. So they are generating um lawsuits automatically completely overloading the law system and we would see the same here. If we automate the processes and then also the input is generated on scale then nobody will be able to check this manually and and we you have to you need a full automatic check on this. We see this in HR.

HR generated CVs are checked by AI agents which which means yeah the entire process is defunct now. Okay. Yeah. No, now I get your point. So that would be closer to the step in the process where it's about um citizen participation because the well the process is started not by individuals but by companies like like grid providers handing in application but the the part with the with the um citizen participation that's actually an issue that if you can you can flood the system with 30,000 automatically generated participation requests whatever um and yes that is uh that's being covered I introduced demos real quick that they they uh they have a solution already in place talking about and we're currently amplifying that solution to to um first of all recognize geni um generated uh input and also to cluster those and to take out well the the um the key uh uh concerns and reply to them uh without flooding the individual or the system with malicious um intrusions and and I I would also add

a point I mean I I would see it a bit different if we have AI anyhow at the input source also companies for example prepare applications also with the help of AI already and then this AI generated content is anyhow flooded to the people to the clerks and if they don't have a a powerful tool at hand with which then they can recognize is it fake is it not fake then I I think uh the the burden is even higher, right? Okay. Um second question there. Please wait for the microphone. Hello. Uh my question is um some public um organizations or governance are not flexible or dynamic. some uh some some of that and um could you give me some highints um about the change management process in this organizations um or challenges and how do you solve that in the past?

What do you mean specifically in in terms of not flexible? Um some have a fear of changes the employees and the decision makers and um um what what are the the solutions in the change management when you plan the projects. Mhm. Yeah. So well it's it's hard to come up with a general reply to that because it's it's a topic for it for its own. But what what I can tell you is that um oh that's a great example. Maybe you've read in the newspaper that the German autoban uh has just awarded an 8-year contract around digitization. Um yeah, some of you maybe have know that we're part of the consortium who were awarded the contract and although we're talking about it's only the autobon uh uh company uh change management is a huge part of that because already well And why that?

Because the client already thought about that up front. So we are getting to a a turning point, I'd say, where it's an integral part of of public tenders to have that topic covered. And if that's not already the case, it's it's well, I think in every consulty's world part of the process to always think that include that topic up front because any technical solution cannot stand on their own feet. But I think the change management has to come from both sides. We of course as technology provider and integrator can help to come up with best practices. But on the customer side, on the public administration side, we also see a need to um show the people that it makes sense for them that it's a relief and it's that they don't have this fear of this technology.

So I think mostly this comes with good use cases that you can show around and say see this really makes our life easier and then I think the adoption increases. So you're showing them how it makes their daily life easier and you come a bit with empathy towards their fear of the change. Yes. Correct. And and also I think nobody ever maybe ever but but nobody nowadays develops an AI solution without involving the users early on. Well, we all do that, don't we? So, we I hope so. Well, usually we we involve the users early on, ask them, well, what what actually is the problem? What do you actually want to automize and how's your how's your work process?

And um so they're they're part of the solution. It's not being they're not on two sides and then they have to have to work with that, but it's usually a joint effort. Yeah. Yeah. coming from a scientific research organization my background I'm I agree to disagree because sometimes it's all about oh this technology can like I can solve the technology problem and not solve for the human you had this beautiful um um logo of Franho um but you're working with the applied science people in the AI department I'm assuming Hinrich Herz institute or one of these um so that is that is um something where you address already these companies who are kind of public sector because Franov is treated like public sector.

Um addressing the problems that the whole public sector has. I mentioned in the introduction of you earlier you worked on vehicles and AI um totally different target group user group. Um how do you address the ethical um considerations um in public sectors like um yeah implement how to implement AI? what you mean specifically with ethical aspects? Uh well, we talked this morning at the um breakfast also with Johan um about um there is GDPR and we should not be afraid of GDPR but we have uh in public sectors um my the details of my um driver's license um my birth certificate and then we don't have that many uh data centers that are in Germany.

So when someone um who owns the data center or is in the country uh where the data center is um then this person can decide okay now I want to have all these details from these other countries because the data centers in my country. Yeah I think this really comes down to the deployment of the technology because I mean we discussed sovereignity here but for example our technology is not sovereign per se if it comes out you have to deploy it in a sovereign way. This means you bring the stack on an on an infrastructure. This could be of course I don't know like the GPU servers of the edit set. It could also be um VPCs that live with Yonos or STEI and uh and then it becomes private and then you make sure that the models that's the beauty of it they don't communicate with the outside world they are a corpose that lift lives in itself and generates output in itself without the help of uh of other tools.

Okay, we don't observe here or include in our observation the MCP protocol or tool calling and so on. This is another word but I think that's how we tackle this ethical problem that the data don't leave uh the secured premises. Okay. Yeah. So you can make sure that it doesn't leave it. Yeah. And and that's why we also make our technology open source because only then you can deploy it on any infrastructure uh that you want. Okay. Thank you so much. You have to hand me the torch because I am allowed to give it to the next speaker before our lunch. And this is your applause. Thank you.

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.