Utilising its full force of glamour, diplomacy, and Parigian style, France rang the wake-up bell for Europe in the race to Artificial Intelligence with the AI Action Summit in Paris. Maybe this came a bit late – but it came with a bang. Emmanuel Macron announced an investment package of 109 billion Euros for AI (and posted deepfake-videos of himself). Leon Wardelmann & Rupprecht Podszun went to Paris to attend the “Business Day” of the AI Action Summit and they tried to find out what role competition and digital regulation play in Europe’s quest for AI leadership. Oh, and they had their Macron moment! Here is their conference debriefing.
By Rupprecht Podszun & Leon Wardelmann
Plug, baby, plug!
Stating the obvious first: We were not admitted to the Grand Palais where Emmanuel Macron met with India’s Prime Minister Nadrenda Modi, US Vice President JD Vance, Ursula von der Leyen and the like to indulge in fancy food and in squabbling over words in a declaration full of buzz-words. The declaration still made the headlines since the US and the UK did not sign it. It calls for AI with a view to “l’intérêt général”? Maybe this was too much Jean-Jacques Rousseau for Vance (who did a bachelor in philosophy at Ohio State before moving on to Yale law).
The geopolitical rift between the USA and Europe is the big topic of the moment, of course, and it was felt at the AI Summit even in the side events. Vance stylised himself as the one embracing tech with no reservations at all. President Macron is also feverishly supporting AI in an exemplary manner. But he still manages to speak of AI and climate change (“plug, baby, plug!”), of the gender gap in AI (women use it less than men), human rights, and AI that is employed not to make a few tech bros ever richer but humankind better off. His address, proclaiming AI as an „innovator for a better life“, also featured a marketing pitch for France as fertile soil for business – highlighting big French strengths: talent (through education) and data centres (through nuclear energy). The AI „moment d’opportunité pour l’humanité“ will be seized with the „Notre-Dame approach“, meaning strategy, timelines, and clear responsibilities – the same kind of esprit that made France rebuild the Paris cathedral within five years (in some other countries, the building permit would probably still hang in fax machines).
In addition to his speech, Macron placed an interview on France 2, where he explained Artificial Intelligence, its risks and opportunities, as well as the government’s role in facilitating AI markets in about half an hour. (If you want to brush up your French vocabulaire numérique this is a great intro!)
But does competition play any role in this? Is digital regulation still a topic in times of AI fever?
Friends of an effective AI toolbox
The best place to look at this question was a side event where even the SCiDA reporters were admitted. Across the city, in the French start-up incubator Station F, led by the shining Roxanne Varza, over 2.500 guests came together for the “Business Day”, including panels and stalls and 1980s-style gambling machines and fancy meeting cubes.
The “Friends of an effective DMA”, i.e. the governments of France, Germany and the Netherlands had put together a morning show on “Winning with AI: How competition fuels innovation”. The people from the economic ministries of the three countries who had once cooperated on DMA legislation put together an impressive line-up. It kicked off with the French Minister of Economics and Finance, Eric Lombard, and closed with Henna Virkkunen, one of the two DMA chiefs in the EU and officially an Executive Vice President of the Commission. In between there was a German state secretary, a video with the Dutch Minister of Economics and two panels with very interesting guests. Daphné Leprince-Ringuet led through the day.
In the first panel, moderated by Pascale Davies from Euronews, Amba Kak from the AI Now Institute hit the ground running with the cutting observation: Not a single (relevant) GenAI start-up does not depend on a partnership with a cloud service provider, obtaining up to 75 % of their financing capital from them. This dependence, combined with dual bottleneck ownership regarding computing power and distribution control, is doomed for market concentration and abusive practices.
Astri van Dyke, a Google lobbyist from Washington DC, made the effort to define a Google open source-policy – she pointed at zillions or so of public papers and a public database. Adam Cohen (one of the many OpenAI people at the Paris event (Sam Altman gave a talk in Station F, too) had difficulties refuting the implications of what Amba Kak had said. He acknowledged the framework of access to inputs and terms of distribution.
French firepower
The organisers had placed Benoît Coeuré in the centre of this debate. He is the President of the Autorité de la Concurrence, the French competition authority, and he expressed his confidence concerning the regulatory way forward. The French watchdog has actually dealt with a real AI case in practice – not just handing out a statement (as basically all competition agencies have done), but a fine: In March 2024, the Autorité fined Google with 250 million Euros for breaching commitments in relation to the press publisher right. In a press release it says:
“With regard to “Bard”, the artificial intelligence service launched by Google in July 2023, the Autorité found in particular that Bard had used content from press agencies and publishers to train its foundation model, without notifying either them or the Autorité. Google subsequently linked the use by its artificial intelligence service of the content concerned to the display of protected content, by failing to propose a technical solution for press agencies and publishers to opt out of the use of their content by Bard without affecting the display of content protected by related rights on other Google services, thus obstructing the ability of press agencies and publishers to negotiate remuneration.”
The agency did not go into the copyright issues directly, but was able to base the decision on the previous competition case regarding the tit-for-tat of Google and publishers. (In a similar case, the Bundeskartellamt had also dealt with the press publisher rights, but without reference to AI, see here.)
Coeuré also pointed at intermediary measures, an instrument that seems to work well in France, less so in the EU or Germany.
It’s a people business
For financing the necessary resources, he took up a suggestion made by Bernhard Kluttig, state secretary in the German Ministry of Economics and Climate Protection. In his intervention, Kluttig had advocated (as his predecessor Sven Giegold, who went back into party politics) to ask digital gatekeepers for “monitoring fees”. Nick Banasevic, representing Microsoft on the second panel, did not seem convinced.
The asymmetry of people in the field is a widespread concern. Adam Cohen: OpenAI employs 2000 people, that is just the size of Google’s legal team. Blanche Savary de Beauregard of Mistral AI, the fat chat of the French AI ecosystem, echoed this. President Coeuré did not hesitate to say that his Autorité handles the package of competition law with just 200 enforcers.
Speaking of talent: Consensus was found on a key bottleneck across all layers of the AI Tech Stack: Talent! Panellists agreed on the necessity for strong European education, be it in academia or the workforce at large. Without talent, there are no innovative companies, simple as that. Here, Astri Van Dyke offered an easy fix: Googlers go on to found start-ups (take AI Summit star Arthur Mensch as an example)! Maybe every European ICT student should just be referred to Google for education.
The traditional clash
The second panel, moderated by SCiDA project leader Rupprecht Podszun (a co-author of this post), focused on the toolkit. Do we have enough enforcement? Do we need more rules? Should the DMA be amended? Asking such questions at the AI Action Summit was not really in the playbook – first, we have to be convinced that AI is the largest thing on earth (and on Mars, Elon, yes). Or, as Euractiv stated in the morning: “Regulation goes out of fashion in Paris”.
The panel offered the dynamics of these fashion moves in full beauty: Nicholas Banasevic, whom many will remember as an anti-Big-Tech-director in DG COMP, the EU competition agency, is now a representative of Microsoft as a Corporate Vice President. (His job change stirred some debate back in the days when he went from enforcement to law firm Gibson Dunn.) Next to him, but with an empty seat in between, sat Max von Thun from the Open Markets Institute, the think tank that is most vocal in advocating the New Brandeis approach. Their positions were quite clear: Von Thun referred to the Gen AI Tech Stack: He sees input factor bottlenecks in chips, compute, proprietary data. Creators of models are dependent, so he said, on partnerships with Big Tech. Von Thun advocates to “reclaim AI from Big Tech” (see his memo here) and calls for a preventative approach with focus on ex ante regulation instead of the “failed” slow orthodox competition methodology.
Banasevic did his best to pacify the snarling tigers of enforcement, referencing the openness in Microsoft partnerships and pointing out, that Microsoft has not been the target of any DMA proceedings (yet?). According to him, there is much competition to be seen. The biggest three cloud providers, so he said, only hold 60 % of the market share – far from the 90 % market share held by one company, as he pointed out (probably thinking of his Google case in DG COMP, but certainly not of the competition law textbook where dominance does not require near-monopoly-market shares). Banasevic, representing a company that currently seems to behave much better than others, looked at Coeuré and Virkkunen who sat in the audience when he reiterated that you really need to look closely into the details.
Unexpected views
The (mild-mannered) Von Thun – Banasevic battle discussion was to be expected. But how did the other panellists position themselves? This came with some surprises. Blanche Savary de Beauregard (General Counsel, Mistral AI) searched for compassion for the fate of smaller not yet super-sized AI companies: unlocking competitive bottlenecks as a start-up while being held up to the same regulatory standards as the major AI players is not an easy task. She referred to regulation in general, like the AI Act.
One man was unexpectedly outspoken: Raphael Auphan, COO of Proton, the privacy-championing e-mail provider, and a serial entrepreneur, did not hold his tongue: He encouraged fierce competition law and DMA enforcement, called for a better funding of authorities and said that Europe was a de facto digital colony of the US. The IT and AI power would enable the new US administration, so Auphan implied, to get whatever they may ask for – Greenland for instance. When the moderator asked the panellists what they would see as the most important next step, Auphan had a clear answer: Europe first. Buy European.
Competition law prof. Podszun – who is not exactly Big Tech apologetic – confessed that such a statement sends shivers down his spine: Isn’t competition closely associated with open markets and a push against national or European champions? The new EU AI Champions Initiative – a case for competition lawyers?
Stephanie Yon-Courtin, Member of the European Parliament and a key driver of digital initiatives there, came to the rescue: Industrial policy and competition law are just two faces of the same coin. She soon went on to call for strong enforcement of the DMA, including support for the “monitoring fee” for gatekeepers. “Don’t hide away”, she said to the Commission.
The European Digital Burger
Yon-Courtin also introduced a culinary metaphor: Her creative description of AI regulation as a Mille feuille, that legendary French pastry. She later transformed the metaphor to a European digital burger, so as to make it easier to digest (sic!) for non-French people. What she meant: The different regulations need to work together and be seen as one big dish, not a piecemeal approach. Those who were hungry for more were disappointed: Yon-Courtin did not clearly speak out in favour of extending the DMA to specific AI services. Enforcement – yes, including AI, where it comes into play, but more regulation does not seem fashionable these days. German state secretary Kluttig had advocated adding AI services to the DMA in his statement.
Asked what she would favour as the next important step, Yon-Courtin also had a clear answer: “Money”. Investment and innovation are fashionable words, competition and regulation less so.
The European Stack
Speaking of a burger reminds us of another idea with several layers advocated these days in Europe: The European tech stack (or EuroStack). Francesca Bria, an innovation economist and a fellow of Mercator Stiftung, called for it. At its core are independent digital public infrastructures that carry European values. The charming idea, set out here for instance, is expensive, but would lead to European tech sovereignty. It is certainly a brick for building something new – the “re-imagination of the digital public space” (as Max Planck researchers call it).
On the day of the AI Action Summit, newspaper Le Monde featured an angry piece by Mariana Mazzucato, the UCL economist. She is a darling of social democrats and regulation-leaning politicians. Mazzucato spoke of “féodalisme numérique” – very much in line with Raphael Auphan. In Mazzucato’s view, AI developers (she names-and-shames OpenAI and Anthropic) are not transparent enough, not declaring how they train their models and what the enormous costs and damages are. This is, in her view similar to the exploitative and predatory behaviour of the big platforms. The words of her and other Big Tech critics are fertile ground for a European response. But our impression in Paris was: This time, the criticism of Big Tech is not fuelling calls for competition and regulation, but for investments into European tech sovereignty. Europe first.
Closing speaker of the day, Henna Virkunnen, sports a title that could not make for a better fit: She is the Executive Vice-President for Tech Sovereignty, Security and Democracy in the European Commission. Digital regulation in the EU is under her leadership. She sat through large parts of the second panel, so that everyone was curious how she would react to calls for more enforcement, a monitoring fee and better regulation. Virkunnen mastered the challenge – by not mentioning the DMA with a single word. Says the Finnish politician: It’s about financing, this will be fostered with the banking and capital markets union. It’s about cutting red tape, and we will deliver on that (35% less bureaucracy for SMEs). It’s about education, research, training the workforce.
Business is Booming – or is it?
Apart from the panels, dozens of companies had built up their stands to showcase their cutting-edge business models. Obviously, the stalls were dominated by the vibrant French scene (SAP being one of the few companies from other countries). Speaking of business models, two shifts became noticeable: First, the shift from big foundational LLMs towards smaller, more focused models. Secondly: There seems to be a shift from private industries towards industries with big public stakeholders, i.e. defence, health care, education and administration. These industries are largely driven by taxpayer’s money. Is it worrying that this seems to be a go-to-place for funding for European AI? Or is it just in line with what Mariana Mazzucato had always said about Google & the likes: They all need public money to get where they are (and therefore we can tax them when they are successful).
The King arrives at last
But all hopes and worries about input markets, bottlenecks or business models quickly faded as rumours spread in Station F that Jupiter was about to show up. And, oui, we had our Macron moment, finally: Tension rose as security and police made their way through the masses. After an hour of waiting or so, Macron finally arrived. (Not sure whether he took the Metro from Champs-Elysée-Clemenceau to Chevaleret with a change at Montparnasse, that should have been quicker than trying to come by car in Paris!) He shook hands and rubbed shoulders with AI business partners and the revered group of AI entrepreneurs. Spotted: Arthur Mensch (CEO and co-founder of Mistral AI), Clara Chappaz (AI minister in the government), Gabriel Hubert (Dust), Antoine Bordes (Helsing) and Véronique Torner (Numeum). If these names do not sound familiar, no worries, we have a steep learning curve. But you would probably have recognised Sam Altman (OpenAI) who even had the red gallic cock attached to his jacket, a sign of French Digital. Maybe some of the US tech bros are looking for presidents that are less difficult to deal with? Macron gave a short, but encouraging speech – “I came to say Bravo to you!”.
The AI Action Summit certainly has brought Artificial Intelligence to the forefront of European media and political discussions. It was not only a French show of force but a signal to European lawmakers and agencies. The focus is on investment and innovation with a European touch. Whether that still includes competition – that is as hard to tell as it is to eat a multi-layered burger stack with French elegance.
Rupprecht Podszun holds the chair for civil law and competition law at Heinrich Heine University Düsseldorf. Leon Wardelmann is a researcher at the chair.
You can subscribe to this blog here – please make sure to check your spam folder for confirmation.
