The decision that there will be a European AI has already been made. The question is who will pay for it and what it will be. Will there be enough political will to reconcile European AI with democracy, human rights? This was discussed at the AI Action Summit in Paris.
This text has been auto-translated from Polish.
The Paris AI Action Summit took place in an atmosphere of forced celebration. Workshops and debates (parts can be viewed) covered climate, energy, labor, collective effort, global competition, energy and the necessity of history. AI can support humanity, give access to knowledge, help find answers to complex problems, but of course it carries the risk of abuse and directing control against people - the concerns put forward by the following speakers were rather subdued, and each doubt-filled sentence was quickly countered with an optimistic one.
Because the decision has already been made. Europe simply has no choice but to enter the technological race, which is starting to resemble the Cold War one a bit, as it proceeds in an atmosphere of tension, confrontation, threats, arrogance and ruthlessness of US corporations in particular.
At the beginning of Donald Trump's term in office, the US announced the Stargate project, for which it intends to spend $500 billion, and the political alliance between technooligarchs and Trump became a reality. In recent weeks, meanwhile, China showcased the project of Chinese startup DeepSeek, which was allegedly going to be much cheaper, but which has little to do with the truth, instead causing confusion. And after all, DeepSeek is not all the Chinese have. I recommend the Techstorie podcast, from which you will learn that some 200 different AI models have already been created in China.
So Europe can have its own AI or be doomed to someone else's. The choice is obvious. If we want to have AI that will be democratically shaped, follow ethical principles, and work for our good ("the common good," "the greater good," "the public good" - various versions of this slogan could be heard in Paris), then it must be ours, European - read: French. The hosts played loudly on national pride, which certainly dampened the enthusiasm of other European countries. However, there is no new ideology behind the French proposals - neither MEGA nor DOGE - only democracy and European values.
Commenting on the differences in approach between the US and Europe, the slogan "the US is innovation, the EU is regulation" is often used. EU reticence, caution and lack of decisiveness have sometimes been the subject of jokes, on a par with fax in Germany or separate taps for hot and cold water in the UK.
But we don't have to choose between full deregulation and total control. We can have fair regulations that don't block development," argued subsequent politicians Petr Pavel, Justin Trudeau and Emmanuel Macron. How? By building our own European AI. A little late to America and China, we still have opportunities, however. Our strengths, Macron said, are energy, of which France has a surplus, and huge innovation potential. The European AI project is also expected to include Canada and India (Narendra Modi is co-hosting the AI Action Summit), which chalks up a new geopolitical deal.
Funding is to be bold and come from a variety of sources. Macron cited as an example the rebuilding of the burned Notre Dame Cathedral, which was accompanied by enthusiasm and widespread generosity, and the result exceeded expectations. Macron wants to generate similar enthusiasm now as well. He announced that France will invest in European AI 109 billion euros, with the United Arab Emirates (50 billion euros), Canada (20 billion euros) and India contributing, and the market, Macron argued, matching that of the US.
Does AI know how to make it into democracy?
How to combine European democratic values with the development of technology, how to use and how to train AI, was devoted to a separate evening event attended by more than a thousand people from 30 countries. Examples of the use of technology against society were presented by Svetlana Tikhanouska: a deepfake prepared by Lukashenko's officials, using her character, in which she was to declare that she was tired and give up. Artificial intelligence can be used to persecute the opposition and ruthlessly control male and female citizens, she warned.
A positive example was given by programmer Audrey Tang, Taiwan's former minister of digitization. She recounted how the government, whose support was seriously waning, decided to use AI to understand what citizens needed. Questions were sent to a very large but randomly selected group of people about specific issues, such as what time school classes should start. However, these were not referendum questions with an expected YES - NO answer, but open-ended questions probing moods, opinions, feelings, emotions, experiences.
The Taiwanese experiment is something we're familiar with from the Greek agora to modern citizen panels - only that with artificial intelligence, it can be done on a larger scale. Not a few, but several hundred thousand minds, experiences, contexts, positions and interests can be involved in working out a given issue. A classic, civilizational achievement dreamed of by practitioners of democracy who want to strengthen it.
When Taiwan's government began using this method, support for it jumped from 10 to 70 percent. Every two heads is not one, and the wisdom of the crowd - the median drawn from a set of responses - is not inferior to expert knowledge. Moreover, the involvement of citizens has emotional value, builds responsibility and solidarity, helps to understand the opinions of others and appreciate the differences, and, above all, trust in the government. A government whose credibility has been tested by the public in this way can dare to make more difficult reforms.
In Europe, this is a story so beautiful that it is almost unbelievable - especially when we relate it to Poland's extremely polarized political scene. A government that wants to solve real problems? A government that listens to citizens? The issue of deregulation would be solved by the Taiwanese government inviting not only entrepreneurs led by Rafal Brzoska to speak out, but also workers, consumers, trade unions - everyone.
After all, one can use AI to survey the opinions of citizens across the EU on the direction of AI itself. This would make it possible to answer people's fears, instead of downplaying them and pushing solutions that are questionable already at the level of declarations. In Paris, assurances were given, of course, that European (i.e. French) AI would respect the free will of citizens, but also the free will of companies. (The event was organized primarily for investors; the only trade union representative there was Oliver Röpke, who simultaneously represented NGOs and civil society.)
This brings to mind the crucial role of political will, which determines how a tool will be used. And political will needs the trust of the public, which was mentioned repeatedly during the conference.
Meanwhile, experience with software that makes decisions for people has so far failed to inspire confidence. One need only recall the accounting software bug, which led hundreds of postal workers in the UK to lose their property, health and families, and sometimes even to suicide. Or the quite fresh and drastic example from the Gaza Strip, where an AI used by Israel called Lavender was supposed to type Hamas leaders, but allowed the killing of hundreds of completely random people.
The control and mobilization tools used by platforms employing couriers or supervising drivers also do not inspire confidence - they do not care about road conditions, traffic jams, the resistance of time and space, which people need a lot of effort to overcome. Also, Amazon has shown an example of using AI to tightly control employees.
The return of the state
What, besides trust, is needed for Europe to defend its AI idea? A strong academy. Mariana Mazzucato, who was present among the panelists, argued that through public investment in the academy, the state not only becomes a hated regulator, but also stimulates development and innovation. And at the same time, it gives itself the opportunity to set conditions and guarantee that the outcome is in the public interest.
It is apparent that the challenges of artificial intelligence - like climate, pandemic, or the threat of war - require that the state be allowed to make market decisions. In America, there has been a takeover of the state by oligarchs; Europe may offer a very different model.
Such a historical necessity will also be emphasized by the panelists in the debate on academia. Representatives of French and German universities unanimously said that a condition for developing and attracting talent is for the state to invest in science. The salaries offered by Big Tech companies are already five times higher at the start than those with which young doctoral students can be tempted by academia. Big tech, in turn, as best exemplified in the US, sucks up public resources and personnel and seeks to monopolize. The result is a withering of innovation, as everyone submits to the prevailing, purely business interpretation. This would not have happened, the debaters pointed out, if the state had included mechanisms to allow fair competition.
As it seems, this is possible precisely in the European Union and precisely - though perhaps paradoxically - thanks to the fact that it is divided into nation-states. This is already slowly happening for the rest: the French company Mistral AI recently showed off its chatbot, which it wittily named Le Chat (such a game of associations, foreigners laugh that in France every word begins with "le", so there is Le Chat, meaning the cat, whose red digital muzzle with ears is the logo of the chatbot). And Polish startup Granary (Speak Leash) is working on White, which is already being trained by the ministry, and the budget for it is said to have no bottom or ceiling. European countries, by sharing software, can develop a variety of AI models tailored to local needs.
This path was presented at a panel on responsible AI by Peter Wang, an AI legend and co-founder of Anaconda, a free software-based data and AI platform. He sat humbly among the audience, long raising his hand to speak. He stood up, introduced himself and urged people not to be afraid, but to create local models, designed for communities, industries, communities. Nobel laureate Daron Acemoglu said similarly: "we will be successful if within two years 80 percent of AI researchers go to their bosses and say: my goal is to work on technology that helps workers, on technology that helps citizens control data and information about themselves."
Capital, corporate interests or power will work against this direction, Acemoglu said, but it is possible. It is worth juxtaposing his statement with the conditions for capitulation delivered in points by J.D. Vance: The EU is to create a friendly and open space for the same model that has monopolized America.
In Poland, the ground is clearly fertile and we are opening ourselves up to the American monopoly for as little as $1 million a year - for five years. Although the Polish presidency is underway, Poland's participation in this European event was negligible.
How to wring out enthusiasm
How to wring out the enthusiasm necessary to give the idea of European AI power when Western society is tired, unmotivated to get involved, and there are so many concerns? When we are losing jobs to AI, and those in power are scaring us with "great replacement" by migrants? "Klarna has revolutionized customer service by deploying an advanced AI assistant that replaces 700 employees." - reported Julia of Infuture.institute in a newsletter.
The report World Economic Forum shows that change will encompass most of the market as early as 2030, that robotization and automation will accelerate, areas of data processing, energy generation, storage and distribution will grow, and that all these trends will exacerbate inequality.
On this, the risks associated with AI do not end. The likely scenario is not only that AI will take our jobs, but that it will start competing with humans for energy and water shortly thereafter. In 2015, a mere 10 years ago, a historic agreement was reached in the same Paris to mobilize governments against climate catastrophe. The passage of the agreement was greeted with great enthusiasm. Today, it's clear that the interests of meat producers and oil companies have won out, and the Last Generation youth protesting against this madness are being put in jail.
We stand at a crossroads. A rational, public-interest view of AI can help us solve many problems, including energy and climate, and yet, as one speaker bitterly stated - AI needs the planet. The question is whether we will follow the logic of solidarity or profit. Then it remains to be seen whether the planet will be enough for all of us.
