CEO & Co-Founder, Mistral AI
The hearing of the head of Mistral AI goes off the rails live!
New video: The Senate welcomes Arthur Mensch, CEO of Mistral AI, for a discussion on artificial intelligence. Among the participants is Communist Senator Fabien Gay, whose remarks did not go unnoticed...
Public Files
Source: Senate, hearing of May 22, 2024
About Arthur Mensch
Arthur Mensch, CEO and co-founder of Mistral AI, has been active in discussions about European technological sovereignty and the strategic importance of open-source artificial intelligence. In a July 2026 panel, Mensch compared the AI economy to the energy sector, arguing that states should consider security of supply, affordability, and sustainability. He stated that open-source AI allows countries to "own" and adapt technology rather than rent it, and he predicted that open-source building blocks will become dominant in AI within five to ten years. In June 2026, Mensch met with Indian Prime Minister Narendra Modi, saying they discussed the importance of countries controlling their AI stack and that he was impressed by Modi's understanding of AI. He also announced a strategic partnership with Airbus at the Paris Air Forum, describing it as a collaboration to embed AI into aerospace systems and data processing, with an emphasis on sovereignty and resilience against third-party decisions. Mensch has also outlined Mistral AI's expansion into manufacturing and infrastructure. In May 2026, he announced deals with Airbus and BMW, describing manufacturing as a "massive market" and stating that Mistral is building models that understand physics and control of real objects. He announced a new data centre in France as part of a roadmap to build 200 megawatts of capacity by 2027 and a gigawatt by 2029. Mensch said Mistral remains on track to reach $1 billion in annual recurring revenue by the end of 2026, and he did not rule out additional fundraising. In a separate interview, he warned that Europe has a "two-year window" to build independent AI infrastructure, citing a €250 billion annual digital services deficit with the US. He argued that building infrastructure in Europe could create a market that attracts chip manufacturers like TSMC or Samsung to set up fabs in the region.
Transcript (10 segments)
A
Arthur Mensch0:00
I would say that regarding labor law, we could give employees more freedom to leave their companies. I think that would largely accelerate things. I'll say it again, it would be very important and very useful. We pay our employees with roughly American salaries, and we still apply French labor law. So one thing we could do, but don't allow ourselves to do, is recruit extremely fast like Americans know how to do, while allowing ourselves to lose 30% of people after two months. In fact, we could do it, but it's not really in the culture. I think French labor law allows it today. So in my opinion, there aren't many changes to make from that point of view. No, really, the issue is to de-protect large companies that want to keep their employees.
F
Fabien1:01
I didn't quite understand about employee rights. I noted one thing: we should allow more rights for employees. So if you could repeat that here a second time, I confess, and even a third time as you leave, it won't hurt anyone.
A
Arthur Mensch1:23
Thank you, Fabien. Regarding employee rights, I'll say it again: we are in favor, generally speaking, startups are in favor of more fluidity in the labor market. That means when we want to recruit someone from a competitor, we'd like to be able to do it in two weeks. We'd like to do it in 10 days, as is done in the United States. But the counterpart is that our own employees can leave, and another counterpart is that it should be less difficult to separate from an employee with whom the fit isn't good. So everyone is well paid. The ideal world for a startup is that everyone is well paid, that you need to be performant to stay in the startup, and that if a startup is going to poach employees from larger competitors, it can do so more easily. So less employer protection and perhaps a little less employee protection on certain subjects when salaries are there, when salaries are high, so we can quickly evaluate if the fit is good and if we made the right bet.
F
Fabien2:31
Yes, Fabien. Not at all on that. I think we need to distinguish, we need to adapt protection to income. I think that's an important issue.
A
Arthur Mensch2:43
No, but wait, don't start, I won't answer on that otherwise we'll go off track.
F
Fabien2:49
No, I'm coming back to what interests me more. So I understood better on the social question, and we disagree, but that's okay, it'll be for next time. A pleasure to see you again to discuss social issues. No, but I think what's important is the union at your place, because at 50 employees, it starts to happen. There aren't 50 people yet in the French company, so we don't have that yet. But I listened carefully to everything you said. I have a point of disagreement or debate, I don't know how to put it. You said artificial intelligence is neutral. I think the exact opposite. The bias with which you build your AI won't be the same as your neighbor's. We are about fifty here. Well, no one is capable of doing what you do. Yes, myself first. Well, there are colleagues much smarter than me, but even the smartest, I won't tell you which one, would be incapable of doing what you do. But if we had the same knowledge, we are fifty, we wouldn't have the same end goal and we wouldn't have the same AI in the end. Because for example, you say afterwards why you did it? It will allow productivity gains. Well, for example, I ask questions: if we have AIs in the future, it will transform jobs but especially work and the meaning of work. Are we doing this collectively to increase productivity or to free up time to do other things? It's a political question that will animate us. So you decide for productivity gains. You can see from where I'm speaking, I would tell you it's rather the second, but that's still a debate. But I'll say it again, when industry implemented Fordism and Taylorism, it put in place tools to increase productivity, it could have done it for other things, but it increased productivity thanks to that. So you see, I'm quite interested in this debate. And the last, very last question. So what drives you? What drives you and why do it?
A
Arthur Mensch5:01
So, on the question of the neutrality of the tool, I think it's never neutral because it carries a societal change. That's the first point. When we talk about neutrality, we're talking more about cultural neutrality, meaning do we allow our users to choose their language, to choose the tone, perhaps to choose the political orientation of an assistant, to choose the policies it will follow and the constraints we impose on it. There are several currents on this. Some companies think they need to impose their own codes and that there are universal codes. And we are more carriers of personalization, and we try to ensure that the tool, even though it carries a societal change in general, can at least be personalized by the users. I think it's useful to come back to what we're really talking about today, which is text generators. So it will transform services. It's not completely immediate. I answered the question previously, but the way it changes production isn't entirely our subject, let's say. What drives us? What drives me is science. I think what drives the co-founders too is being able to push science and to do it independently, without being vassalized to the United States, without being vassalized to existing companies that had enormous success 25 years ago and use that success to consolidate their dominant position. For us, it's important to perpetuate these cycles of innovation and to launch innovative companies that will bring different things than what large existing companies can bring.
F
Fabien7:04
No, on the change in production, it's still a debate and it's also your debate. I'll take a field I know a bit, the world of the press, it will have impacts and it already has impacts today, but they aren't neutral. I mean, today, for example, Elon Musk just explained he's going to invest billions to create a website that will be an information site where articles will be generated exclusively by AI, with everything that comes with it, fake news, destabilization of other countries. It's already the case today. So for example, we can use AI to assist in certain jobs in the press, but tomorrow, are we collectively in agreement that articles, whatever they are, should be produced by AI and no longer by the human brain? If you tell me no. And so it already has a concrete impact because the reality is that AI today is already generating articles that are relayed powerfully by algorithms on social networks, and that's going to be on a scale of 10 years more than the real. That will be the concrete reality. So when you're in control of this, saying we put all this at our disposal but we keep at arm's length the question of what it will transform in production, you see...
A
Arthur Mensch8:35
On the subject of journalists, it really needs to be seen as a tool. That is, what AI doesn't know how to do is be on the ground. It's doing interviews, it's going to see what's happening. What it can do is facilitate investigation. When you receive the Panama Papers, it's not bad to have an AI at hand to go find who had their hand in the cookie jar. And so I think we really work with several newspapers on these questions because I think it's really a tool for making better information. Afterwards, it's also a tool for making disinformation. So that means we need more moderation on social networks, for example. And on the question, but in general, the question we need to ask is that this transformation and these tools are arriving. And the question is, is there someone here? Is there someone on the territory in Europe trying to build these tools to contribute to the debate? Because otherwise, the tools aren't ours and we'll suffer it all the more.
N
Narrator9:37
And what do you think? Tell us in the comments. Thank you for watching this video.