Prime Minister Kyriakos Mitsotakis participated in a conversation with Arthur Mensch, co-founder and Chief Executive Officer of Mistral AI, in the context of the GenAI Summit at the Stavros Niarchos Foundation Cultural Center. The discussion was moderated by journalist and member of the GenAI Summit Organizing Committee, Nikos Efstathiou. The Prime Minister’s remarks follow:
In his introductory remarks, when asked about the memorandum of understanding between the Greek government and Mistral AI signed today, the Prime Minister stated:
Well, first of all, it’s a privilege to be on this stage. Again, congratulations for a team that is not very well-versed in event management. You’ve done an amazing job in putting together so much talent to Greece. And thanks to Arthur for being here and sharing the stage with me.
As we’ve developed our AI strategy over the past years, our goal has always been to team up with leading AI companies to establish mutually beneficial partnerships. When the opportunity presented itself to sign a memorandum of understanding with Mistral, this was too good an opportunity to pass by. I’m very happy that we’ve been able to conclude this agreement. Mistral is a leader in foundational AI models.
As you pointed out, Arthur and his team have proven that you can actually build these types as a model in Europe, incorporating European sensitivities and European values, but also being much more aware of what I may term as probably more “European regulatory sensitivities” when it comes to AI.
I was particularly struck, when the opportunity was first presented to us, with the sensitivity that Μistral has shown towards preserving linguistic and cultural heritage. Greek is a language that has been spoken continuously over thousands of years. The ability to understand the nuances of the language was particularly important to us.
But I was also fascinated by the work that Μistral has done with public administrations, not just in Europe, but also in Asia. One of our big theses is to be able to use AI to improve the efficiency of government, both in terms of customer-facing applications, but also the internal process of how the bureaucracy works. The fact that they’ve worked on this front with numerous countries was very interesting to us.
Of course, the opportunity also that Mistral will provide up to 30, maybe who knows, even more Greek STEM graduates to have placements with Mistral is also an indication of how much importance I place in ensuring that our top graduates have a professional experience with the leading AI companies.
I think what makes Greece rather unique, in spite of our size, is the disproportionately big representation of top AI talent globally. This is something we intend to leverage.
But of course, we also need to think in terms of the next generation of AI leaders and ensuring that our graduates work and train with the best is a top priority for me.
When asked about the prioritization of areas of cooperation between the Greek Government and Mistral AI, the Prime Minister noted:
I’m always very interested in what Arthur said: what are the really applicable solutions that can increase the productivity of our civil service, and how can we make the interaction between citizens and businesses without unnecessary friction.
If you look at the history of what we have done in terms of our digital transformation over the past years, we started with the basics. First of all, I think it is very important that in 2019, we set up a Ministry of Digital Transformation. We were one of the first European countries to actually do that, and we endowed it with horizontal responsibilities to work with other ministries.
But also we made sure that, in terms of the control of the data sets that the government had, that they would have an important role to play. This is becoming more and more important in today’s landscape, when everyone is looking for well-curated data. The fact that we’re also a relatively centralised country facilitates this exercise.
Look at healthcare, for example. In our case, for the companies that are interested in partnering with us, there is relatively well-curated healthcare data that is related to a population of 10 million, but you only have to talk to two ministries, essentially, rather than talking to regions and trusts and diversified hospitals. I think this makes the cooperation between companies and the state also easier.
But when I look at also the partnership with Mistral, the one area that stands out, and we’ve discussed this, for example, is justice, how much AI technology can help in simplifying a lot of procedural manual work that takes place in the court to help our judges and the support staff focus really on what is important, and that is quality judicial decisions.
Of course, the topic where I think AI is going to be truly transformative, and we’ve also established partnership with other leading AI companies, is education. I think we are at the beginning of what is a massive transformation of how we address the needs of our children.
Right now, we have a factory model approach where essentially we teach our children more or less at the same speed. But if you look at the possibility to deliver customised support to students, at a time when we already recognise that our students are using AI models for their own purpose, what does this mean for the whole educational process?
I think this is a global challenge. Of course, we in Europe need to develop our own strategy in terms of how we integrate these momentous developments in our educational system.
Regarding the impact of artificial intelligence on public administration, following the progress achieved by Greece in recent years, Kyriakos Mitsotakis stressed:
First of all, I think it’s important to educate our civil servants about the potential of AI tools as a useful assistant to improve their productivity. We’ve engaged in these types of programmes.
The more we look into the details of processes and on specific projects, especially of repetitive projects that entail heavy bureaucratic work, the more we can come up with tailor-made solutions to help our civil servants do the job faster and better.
For example, one of the first pilots we ran was with our notary services in terms of an AI model quickly reviewing a contract and essentially reducing the time of review from 2 hours to 10 minutes, because most of the work was done and only some final legal control needed to be applied by humans.
Every process can be different and every challenge can be different. But it is about infusing our civil service with a different mentality. It is also about providing cost-efficient solutions, sometimes using open-source models that can be easily customised to address specific needs.
I was very recently in Singapore, and I had a very interesting and long discussion with the Prime Minister, who described their own experience, and they are an incredibly advanced country in terms of public administration.
Singapore decided that rather than sort of outsourcing this knowledge to outside contractors or technology providers, they could build their own in-house technology capabilities to work with models and contractors, but to retain the intellectual property of the solutions that they develop. I mean we’re trying to do something, I wouldn’t say similar, because Singapore has thousands of engineers working for the government, just to give you an indication. Of course, as you know, Singapore pays very well. They do pay competitive salaries, which is something which is very difficult for us.
But we’re building our own small in-house team of engineers that is able to look at these problems and come up with customizable and relatively cheap solutions, again, looking at specific problems that need to be addressed quickly. The more we develop this positive momentum, I think the more interest we will have from other parts of the civil service to come to us with problems that then we can solve.
Of course, by looking at best practices from other countries – and that’s why these partnerships are so important – we can get very interesting ideas. We discussed with Arthur this idea of our labour placement agency, having agents that do better matching between the request of employers and what is available in the database of the labour employment agency in terms of talent that can be used or that we need to place.
So, this is a lot about flexibility. It’s the same way we also build our gov.gr model. The more we build the trust within the administration, the more agencies will come to you and say, “look, I also have to benefit”. This is how gov.gr started. In the beginning, there was a lot of skepticism, but at some point, the relevant ministries realised that there was a political benefit to them if they digitised part of their own processes. They also had an interest in joining this effort, because they understood that they would also gain some political benefit from doing so.
Sο a lot is about education. A lot is about examples, specific examples of what does it mean to actually use this technology to improve productivity, and of course, changing the mindset and not viewing, especially in the civil service, where job loss is not really a threat. This is a big distinction between a private company and the civil service. You don’t risk being replaced by an AI agent. That is not necessarily the case when you talk about private companies. I think it’s easier to build this level of trust, that AI can be a true partner rather than a threat in terms of helping our civil servants to improve their productivity.
When asked about the use of artificial intelligence applications in public administration and citizens’ trust in the adoption of artificial intelligence, Kyriakos Mitsotakis replied:
Well, I think there’s no obvious answer to the question that you asked. I think the speed with which citizens are using AI tools is just incredible. It seems to me that right now, at least this is what I’ve seen from some research, that they may place more trust in the answers that they can get by asking an AI model than the trust that they have in terms of getting information from the traditional media or another source. This, of course, places a lot of responsibility on the development of these models, because at the end of the day, they’re also receiving very specific input in order to come up with the answers that they provide.
I think we’re faced with a broader problem, and that is an industry that is moving at break neck speed. Governments, not sometimes, I would say almost always, not exactly understanding the speed with which the technology is changing.
And then you’re caught between two different possible priorities. The first is move us quickly and make sure that you don’t sort of restrict this emerging industry that I would argue is probably the American approach of, I would call it “extreme deregulation”. And the European approach which says, “Okay, this is something that we need to approach from the perspective of regulating in a smart way without inhibiting innovation”.
Of course, I’d be very interested to also hear Arthur’s view on the regulatory attempts of the European Union in terms of the AI Act.
But it is clear to me that finding this balance is not obvious. I don’t think the solution is complete deregulation. On the other hand, we need to ensure that we maintain the dynamism of the European AI startup ecosystem, which is also very much demonstrated in this gathering today, and that we don’t send a signal that Europe is a continent that essentially makes it difficult for these companies to scale up and thus sending a signal to all those sort of talented young people or young engineers that it’s probably better to go work for a big US company, rather than working for a company such as Mistral or a smaller, in this case, Greek startup.
So, it’s not easy to find an answer to these challenges. In my mind, there are areas where one clearly would require a more regulatory approach.
Two topics which I always raise in my public comments, which are of great concern to me. The first is the mental health of children and teenagers, which is also related to the way our kids use, for example, AI for school, which is, again, a real challenge because it is happening as we speak. And the second topic relates to the challenge of maintaining our democratic process intact and immune from threats of AI agents that can easily play havoc with the flow of information and thus directly interfere in our democratic process.
There’s a question of whether one can use “wise” AI tools as a response to these challenges. Let’s say if one could have, I don’t know, some sort of “wise” AI chatbot that could debunk fake news, for example. I mean, one can think of constructive ways of using technology to build trust, because if at some point there is so much information flooding the public sphere, people will end up believing nothing. This, I think, is profoundly detrimental to our democratic process. It’s probably what those who oppose the European way of life would like to see happening.
When asked whether he considers it important for Europe to have its own strong ecosystem in the field of artificial intelligence that reflects its values, and what role he believes smaller European countries can play in supporting or strengthening European companies, the Prime Minister noted:
First of all, the answer is clearly, yes. I think Mistral has proven that one can develop a global leader in the foundation of AI models in Europe. I think for the reasons that we discussed previously: a better understanding of the European way of life, a greater sensitivity with this balance between innovation and regulation, and plainly the fact that many Europeans may feel more comfortable working for a European company, even for practical reasons – they don’t have to move to Silicon Valley, but can stay in Europe.
This, I think, supports the argument that, of course, European players need to have a special role, and we also need to listen to them when it comes to adjustments to our regulatory framework. We’re not in competition with industry. We may have our own priorities, which we discuss with industry, and hopefully, we can convince industry to also adapt to our sensitivities.
For example, I’m coming back to the question of mental health of kids and teenagers. Since we started raising the question with many American providers of applications which are sensitive in terms of interacting with teenagers, we also saw momentum on their side to put in place controls and requirements, first of all, acknowledging the problem. It may not go as far enough as we would like, but at least it shows that this constructive dialogue can actually be productive.
Of course, we should not forget that beyond the big players, there’s a whole layer of startup companies primarily focused on AI applications emerging in Europe. I’m very optimistic with what I see in Greece now. The tech and AI startup system has truly exploded. The number of venture capital funds active in Greece is something I would never have imagined a decade ago.
To be able to provide these companies with a European path is very important. Because at the end of the day, what I would not necessarily want is for the big American tech companies to scan Europe and to immediately acquire these companies at the valuation, which may be quite meaningful for the founders, but which may still leave money on the table in terms of offering these companies the ability to further grow in Europe.
Of course, in that sense, two priorities from our point of view as someone who’s sitting at the European Council: the first is European capital markets union, what we call Savings and Investment Union, the ability to actually raise capital in Europe rather than going to the US.
And what we’ve been calling the “28th regime” for startups companies, the ability to seamlessly do business in all European countries without legal and bureaucratic constraints. This is something that has been raised rather persistently by those who have studied the single market.
We have a single market, but in digital services, it does not work as it should. If you look at it, essentially, it’s an internal tariff that we put on our single market. We could benefit so much more by streamlining procedures and ensuring that a tech company in Greece could operate in all EU member states without additional regulatory concerns.
I think these would be two priorities that will clearly strengthen our position in the technology space and would allow for more European companies to emerge, to develop, to grow in Europe without risking being acquired at an early stage by the big American hyperscalers.
Responding briefly to questions, the Prime Minister stated:
What’s the one word that you think should guide Europe’s AI future? “I would say ambition and responsibility”.
What do you think is the sector where AI will transform Greece the fastest? “I would say the public administration, but in general, I would say health care”.
What is one thing you think Europe must do immediately to stay competitive: “Pass the digital omnibus. I’m saying this because we all talk about deregulation, and theoretically, this should be the lowest hanging fruit, not just in terms of digital technology, but also in terms of making the Green Deal more respectful of the regulatory burden we’ve put on smaller companies. But for some reason, we’ve had difficulties moving from the intention to the end result”.
What is one Greek word that you like to see AI models like Mistral’s, learn to use correctly: “Philotimo, if you can translate that”.
What do you think success will look like, in two years, for the partnership that the Greek Government and Mistral AI signed this morning: “Have one flagship project, let’s say in justice, that you can say in two years: Greece did it first.”
