At the Parisian show dedicated to new technologies, artificial intelligence is the subject of all companies, as a lever for solutions or a trigger for projects. Everyone wants to ride the new wave, but in Europe, the AI Act intends to regulate this popular area.
6 mins
Difficult to miss. In the aisles of the VivaTech show, which is held at Porte de Versailles, in Paris, AI is among all the technological solutions presented. If we do not always understand its usefulness, according to the exhibitors, it is never futile. It has even become a professional commitment, for Phillippe Lemaire, founder of Safebrain: “ I started as a cybersecurity expert and when ChatGPT was released in November 2022, it was like an epiphany for me. I said to myself two things: the first is that it would really change the world and the way people work – in 20 years, I have never been surprised like that in front of a piece of technology – and then, I said to myself that It was imperative to be able to use generative AI while preserving the sovereignty of our data, both for the country and for businesses. »
Data is the major issue behind this new technological leap. For what ? “ Because AI can be incredibly relevant and give advice that we cannot do without, but for that you have to give it data and a lot of data that I call “intimate data”, continues Phillippe Lemaire. We talk a lot about personal data with AI, in fact, we give it important and precise data that can be considered intimate and that must be secured. You have to be careful about what you transmit and that’s why we developed the solution SafeBrain, a ChatGPT for businesses. »
“ The AI Act reminds me of the GDPR law »
The AI Act, definitively approved at the beginning of the week, promises to classify AI according to their level of risk: unacceptable, high, limited and minimal. Future regulatory measures will vary accordingly, with a desire for transparency: users must be informed when they interact with an AI.
But other implications are to be expected for European companies: increased compliance costs, even administrative complexity, which would slow down innovation, with the consequence of a competitive advantage for non-European companies – mainly American companies. This is an argument pointed out by many players in the field, including this business leader we met at the show: “ The AI Act reminds me of the GDPR law. In the United States, I had a company in Silicon Valley that was doing well and we knew that Europe was working on this legislation. There were then consultants [de grandes sociétés américaines, dont Meta] who visited all the Silicon Valley start-ups to find out what we saw as good and bad in the future GDPR and what it would take to ensure that it did not penalize American companies too much. [Ces lobbyistes] have made their shopping list with all the startupers and big American companies. ” He smiles : “ And all of our requests were heard by Europe, for the benefit of American companies, and we knew very well that it was going to be to the detriment of Europeans. Well, the AI Act is exactly the same. »
However, this is not quite the direction in which the IA Act is committed, which aims to encourage innovation and support small and medium-sized enterprises (SMEs) by offering them exemptions or facilities to comply with the rules. . In addition, it promotes cooperation with third countries and international organizations to harmonize standards and practices around AI (the summary of the draft regulation can be read here).
“ Europe absolutely must have sovereign AI »
At the house of SafeBrain, we are realistically observing the arrival of AI. “ It is obvious that this is a new Wild West, we must be aware that the Americans use [nos] data, we have known since Edward Snowden, it is a major subject. We are entering politics, we are no longer in technology, so do we consider that we must protect the data and not give it to the Americans, or do we consider that they are allies and whatever? asks Phillippe Lemaire, who has his own idea : “ The next AI models will only have as a training source the information that we will give them with our accounts [sociaux]. We need original data and this original data will be our interactions with artificial intelligence. We must not be candid, Europe absolutely must have sovereign AI so that companies and public organizations can protect themselves. »
The opinion is even sharper from our French entrepreneur: “ The big Americans like OpenAI and Google are working hard on AI and Apple will perhaps join the race, but in France, we have Mistral which is working very well and rather than regulating, rather than spending money in an IA Act, if we gave this money to Mistral to make him a champion, wouldn’t that be better? »
Taking the example of the GDPR, he evokes the fate of OVH, a “cloud computing” company: “ We had an extraordinary company, which still exists, but we did not work together to put all our money in the same basket, to make OVH a European champion, a French one at that. We are beside our pumps, we work for the Americans. » Recently, on this same subject, Michel Paulin, CEO of OVH France, rightly regretted “ a rush on AI that seems rushed, so “ that we are barely understanding what is happening in the field. »
While the European AI Act, which will come into force in 2026, is the subject of debate on the VivaTech stands, it has not cooled the atmosphere, nor has it overshadowed the big announcements at the show. A new French generative artificial intelligence start-up called H was launched on Tuesday, with a fundraising of 203 million euros from large investors including Bernard Arnault, Xavier Niel and Amazon.