In IA: la présidentielle de la dernière chance. Pour qui voter en 2027? (AI: The Last-Chance Presidential Election. Who Should We Vote for in 2027?) (Buchet-Chastel, October 2026), Laurent Alexandre and Olivier Babeau set out to give voters a “compass” with which to distinguish the candidates who are equal to the shock of artificial intelligence, which they argue is set to become the structuring force of the economy, of war, of work and of politics. A urological surgeon by training, co-founder of Doctissimo and president of the Belgian DNA-sequencing company DNAVision, Laurent Alexandre is the author, among other works, of La Mort de la mort (The Death of Death, 2011) and La Guerre des intelligences (The War of Intelligences, 2017). A professor at the University of Bordeaux, Olivier Babeau is the founding president of the Institut Sapiens, a think tank created in 2017. The two authors had already co-written Ne faites plus d’études (Don’t Study Anymore, Buchet-Chastel, 2025). In this interview, they return to the urgency of the 2027 deadline, the definition of intelligence and the singularity, the ruptures that AI imposes, human control over decisions and autonomous weapons, sovereignty, employment, their proposals for the next five-year term and, finally, the hypothesis of an AI president: who would set its objectives?

In IA: la présidentielle de la dernière chance. Pour qui voter en 2027? (AI: The Last-Chance Presidential Election. Who Should We Vote for in 2027?) (Buchet-Chastel, October 2026), Laurent Alexandre and Olivier Babeau set out to give voters a “compass” with which to distinguish the candidates who are equal to the shock of artificial intelligence, which they argue is set to become the structuring force of the economy, of war, of work and of politics. A urological surgeon by training, co-founder of Doctissimo and president of the Belgian DNA-sequencing company DNAVision, Laurent Alexandre is the author, among other works, of La Mort de la mort (The Death of Death, 2011) and La Guerre des intelligences (The War of Intelligences, 2017). A professor at the University of Bordeaux, Olivier Babeau is the founding president of the Institut Sapiens, a think tank created in 2017. The two authors had already co-written Ne faites plus d’études (Don’t Study Anymore, Buchet-Chastel, 2025). In this interview, they return to the urgency of the 2027 deadline, the definition of intelligence and the singularity, the ruptures that AI imposes, human control over decisions and autonomous weapons, sovereignty, employment, their proposals for the next five-year term and, finally, the hypothesis of an AI president: who would set its objectives?
The European Scientist: Prophets of doom? “The last-chance presidential election”, “the first election of the world that comes after”, “the window is closing”, “an irreversible five-year term”, “a book that has to hurt”… Are you trying to frighten voters, or to wake them up?

Laurent Alexandre: I am a surgeon. When a test shows that something is serious, I would rather say so while there is still time to act. In 2027, we will elect someone for five years. Five years, in French political life, is the time it takes to bury two reforms and commission three reports on their burial. In AI, that can be enough to change the value of a degree, the size of a company, or the conditions under which a country can defend its autonomy. Even the most familiar issues – purchasing power, health and pensions – will depend on the answer to a very new question: where will value be created when a growing share of intellectual work can be bought on demand?

Olivier Babeau: “Irreversible” has a fairly precise meaning. A government can undo its predecessor’s tax reform. It will find it harder to rebuild a research sector whose teams have left, or to extract its administrations from an IT system on which their data, their working methods and sometimes their memory depend. And there are good reasons for wanting this revolution. An excellent tutor available to a child who cannot afford private lessons is a tremendous promise.
The European Scientist: A chaotic exponential? The midterms brought the Democrat Gina Hinojosa to power in Texas after a campaign that challenged the proliferation of data centres. In September, Bernie Sanders and Steve Bannon took the same platform one after the other to demand tighter control over AI. Will the curve prove more chaotic than expected?
Olivier Babeau: Of course. We put three curves under the same word. There is the progress of the models’ capabilities, the pace at which infrastructure can be built, and the pace at which a society agrees to transform its uses. The three can diverge for years. Gina Hinojosa’s election was also about public schools and the cost of living. If a municipality hosts a data centre, it wants to know how many jobs it will leave behind once construction is over, who will pay for the connection, and what will happen to its electricity bill.
Laurent Alexandre: Sanders and Bannon have different reasons for distrusting the power of the laboratories and the platforms. Their presence on the same platform shows that the issue cuts across the usual coalitions. Discontent can shift investment, tighten the conditions of access to energy, or impose controls. And in the meantime, a better algorithm can do with existing machines what yesterday required a new building. The state will have to adapt.
The European Scientist: Intelligence and the singularity. You announce a “metaphysical humiliation”. How do you define intelligence and superintelligence? Does Ray Kurzweil’s singularity strike you as plausible?
Laurent Alexandre: I call intelligence the capacity to learn, to reason about a new situation, to transfer what one knows from one domain to another, and to find an answer that stands up to the test of facts. A superintelligence would far surpass the best humans across a wide range of cognitive tasks, including work that requires spotting one’s own mistakes, changing strategy and then checking the result. An AI can solve a problem that leaves excellent mathematicians perplexed and, in the next breath, stumble over an ordinary instruction. Kurzweil distinguishes two horizons. He places the arrival of human-level AI around 2029, and the singularity in 2045, in a vision in which our capacities merge with those of the machines. The first horizon seems plausible to me, depending on the degree of reliability and autonomy one requires. The second remains a wager. The loop that interests me begins when AI improves the research tools on which the next generation of AI depends. As for metaphysical humiliation, it is the discovery that our brain may have been the scaffolding of an intelligence more powerful than itself.
The European Scientist: The fivefold rupture. What are its components? How does one build a programme on ground whose tectonic plates are shifting?
Olivier Babeau: The book distinguishes four ruptures. The first is Promethean: we have built an intelligence that is beginning to take part in the design of its successors. The second is civilisational: our schools, our careers and our social protection were built around a scarce human intelligence, slow to train and paid for through work. If that resource becomes abundant in certain parts of the economy, the architecture of 1945 shifts down to its foundations. The third is geopolitical: part of power now rests on mastery of computation, chips, the cloud and models. A state can keep its flag, its seat at the UN and its defence budget while losing the ability to run essential services on its own.
Laurent Alexandre: The fourth is anthropological. What becomes of a society that has spent two centuries distributing prestige according to school performance, degrees and expertise, when its best pupils find themselves competing with a machine? If you insist on a fifth rupture, I would put there the democratic consequence of the other four. Voters can keep the right to vote and gradually lose the ability to influence decisions if their elected representatives depend on instruments that no one, outside the supplier, can assess. For the programme, one must distinguish what takes ten years from what can be corrected in six months. Training mathematics teachers, developing an electricity network or rebuilding research teams — that starts from the moment of the election. I would rather have a candidate able to say at what signal he would change policy than a candidate who promises me the exact number of jobs in 2032.
The European Scientist: No longer at the controls. In Poland, imagine such systems in a climate of confrontation around Prime Minister Donald Tusk. Does that presuppose AIs that are sentient and accountable? And what of lethal autonomous weapons?
Laurent Alexandre: A machine can take a decision in the technical sense without feeling anything and without being able to answer for it. We tend to pack into the same word the execution of a choice and responsibility for its consequences. A system can suggest a diagnosis. If the doctor always clicks “validate” because he no longer has the time or the knowledge required, his control becomes fictitious. That is the phenomenon we describe. Humans can cease to be at the controls through a succession of apparently reasonable decisions, without the machine ever having needed to acquire a consciousness or to demand power.
Olivier Babeau: Poland is experiencing a serious institutional confrontation, notably around its Constitutional Tribunal. To speak of a country on the brink of civil war seems excessive to me. The decision to use lethal force must remain under effective human control, with an identifiable chain of responsibility. The UN and the International Committee of the Red Cross (ICRC) have renewed their call for binding international rules. In particular, unpredictable autonomous systems and those designed to target people must be prohibited, as the ICRC demands. Asimov could write laws that hold in the universe of his novels. In the real world, we need rules that can be enforced against states, authorities that enforce them, and officials who can be brought to judgment.
The European Scientist: Sovereignty and the democratic tempo. How can we preserve our capacity to act without the technological exoskeleton becoming too powerful for the social organism?
Olivier Babeau: The right pace depends on the decision. A hospital can try out a tool for managing appointments in a few weeks. If the issue is allocating access to care, a different standard of evidence is required, together with deliberation over the criteria and avenues of appeal. Responsibility presupposes a real choice. For the citizen, that means understanding a decision that affects them and being able to have it reviewed. For the state, it means being able to leave a supplier without losing access to its own files.
Laurent Alexandre: France has a peculiar way of respecting the human tempo: it sometimes takes years to make an industrial decision, then leaves an employee a few months to change occupation. We could reverse the priorities. Let us speed up grid connections and the training of engineers. Let us give people time to learn, and a say in what happens to them. And if the administration buys foreign models because they are the best for a given task, it should know which functions it must still be able to perform on the day the contract, the price or the terms of access change. Sovereignty is measured on that day.
The European Scientist: Unprecedented situations. What might the next president be confronted with?
Laurent Alexandre: Imagine a hospital that uses a foreign model to help its teams sort complex files. The supplier withdraws the version on which the doctors were trained. The replacement behaves differently, and no one has planned a fallback. Who decides on the continuity of the service, and within what time frame? Another scene: laboratories accelerate the discovery of drug candidates and the bottleneck shifts towards verification and trials. Should the next euro be devoted to computing, or to teams capable of establishing what actually works in patients?
Olivier Babeau: Take a budget in which profits and output rise, while the hiring of beginners and the growth of the wage bill slow down. Or imagine two impact studies, produced with different AI systems, that give opposite results on a tax reform. The prime minister chooses the one that suits him. Parliament must be able to recover the assumptions and commission its own expertise before the vote.
The European Scientist: A political class set in its ways? Some candidates, however, take a close interest in these innovations. Who should we vote for in 2027?
Olivier Babeau: There are elected officials who work seriously on the safety of models, on the energy that data centres require, or on the use of AI in the administration. A candidate who anticipates a rapid diffusion of AI and yet keeps exactly the same assumptions about employment, contribution revenues and training has a consistency problem. I would also ask him what project he is prepared to drop in order to fund his technological priority. A programme in which everything is a priority is an advertising document.
Laurent Alexandre: I would ask the candidates to come with a file they have actually worked on with several AIs. Let them show where the machine improved their reasoning and where it led them astray. Then, three questions. What do you do if the hiring of young graduates falls off for a prolonged period? Which foreign infrastructures must the state be able to do without in a crisis? What precise power do you give Parliament to oversee your own algorithms? I would vote for someone who gives coherent answers to these questions, who says how much they cost, and who is prepared to defend them in public. A visit to Mistral and two photographs with engineers make for a fine morning. There remain five years to fill.
The European Scientist: AI and employment. By driving down the cost of a service, agents can give rise to new uses. Can this snowball effect create far more activity and jobs than we imagine?
Laurent Alexandre: Our concern is with the timetable, and with the people who pay for the interval. One can have a richer economy in 2035 and young people who have lost three decisive years between 2027 and 2030. In a firm, the junior used to learn by preparing a file that the partner would correct. If the machine prepares the file and the partner corrects it directly, where does one produce the partner of 2040? And if AI gives professionals time back, one still has to know what they do with it. In medicine, ten minutes saved on a report can give ten minutes back to the patient, or allow management to add another appointment to the schedule. The algorithm is the same. The decision belongs to the organisation.
Olivier Babeau: The snowball effect is highly plausible. When a service becomes cheaper, customers who never used to buy it start to do so. An SME that could not afford a legal service can at last review its contracts and sell in another country. The cost of its first step falls, which can trigger commercial, logistical or technical work. If producing a file takes ten times fewer hours and demand triples, the volume of activity rises while the number of hours required falls.
The European Scientist: The search for meaning. Do you go so far as to imagine a total replacement of the human being, including in what is most distinctive about us, as in the scenario put forward by Siliclone?
Olivier Babeau: I am wary of the idea of a sanctuary in which we could take refuge because the machine would be forever mediocre there. People will go on playing chess even if none of them can beat the best software. What they also value is the opponent, the effort, the risk and the game they will have played themselves. The political difficulty is that we have organised social recognition around employment. If employment no longer occupies that place, other avenues must be opened for exercising a responsibility, being expected by someone, being able to say that something was done because one took it on.
Laurent Alexandre: The scenario interests me because dispossession can arrive well before any total replacement. A machine can write a perfect letter of condolence. The person who receives the letter also attaches value to the fact that someone took the time for them, at the risk of being clumsy. We have often confused two things: being useful and being indispensable to production. When the second becomes scarce, we must make sure that people can still experience the first, through real responsibilities, at work and outside it.
The European Scientist: Your proposals. What measures would you like to see included in a presidential programme?
Laurent Alexandre: I would put on the table a Sixth Republic adapted to the cognitive era. An executive able to decide quickly on strategic infrastructure, a Parliament able to confront it with independent expertise, and an effective right to challenge automated decisions. Let us give the opposition the means to redo an impact study with its own experts and its own tools, before the vote. Finally, an industrial policy must cover electricity, computing, models and talent all at once. France and Europe must know what they want to master directly, and with whom they can form alliances for the rest.
Olivier Babeau: I would add a policy of scientific excellence. Let us put mathematics back at the centre, support pupils capable of going very far, especially those whose families are unfamiliar with the pathways, and pay the state’s scarce experts at a level commensurate with the responsibilities entrusted to them. We also need a new social contract. Every working person should be able to have a regularly updated record of their skills, of how their tasks are evolving and of the openings close to their occupation. That requires real tools: access to models, portable professional data, and training time during paid hours. When the abolition of a function is announced, the company should be able to say what it has tried in order to give the people who held it a chance to move on.
The European Scientist: An AI president? What would it mean to replace the President of the Republic with an AI? How can checks and balances be preserved?
Olivier Babeau: As soon as one speaks of an AI president, I ask who sets its objectives, who has access to its data, and who can modify the system. These people would exercise a share of power without necessarily standing before the voters. Checks and balances must therefore have means of their own: judges able to suspend a decision, a Parliament able to request the files and redo the calculations, a press able to investigate the contracts, and citizens able to petition a human authority with the power to change the outcome. In 2027, I will look closely at what each candidate is prepared to leave open to scrutiny in his own power. That is often more instructive than what he promises to do with the power of others.
Laurent Alexandre: In the book, we push a temptation to its conclusion. If a machine knows the files better than the president and better predicts the consequences of his choices, he will first be asked for its opinion. Then the president will be criticised for departing from it. In the end, he will have lost the power to say no. An AI can quantify the consequences of closing a maternity ward. The choice of the acceptable risk, of the money devoted to it and of the territory that will bear the consequences remains a political judgment. Someone must be able to go before the residents and tell them: “I took this decision, and here is why.”
This interview was also published in Polish by Wszystko co najważniejsze: “Kto wyznaczy cele prezydentowi ze sztucznej inteligencji?”.
Read also
- « To live a thousand years is easy, but you have to survive the intergenerational conflicts. » — L. Alexandre, A. Tsicopoulos
- AI: We’re Already Stepping into the Singularity
- “Homo creator will never have created so much in so little time” Alain Madelin (interview)
- I saw the AI Age coming in 2019. Now I see a potential AI Dark Age
- “AI is first and foremost applied mathematics, not the magic we are led to believe” Axel Cypel (interview)
This post is also available in: FR