Chapter 28
Monday, November 2, 2026
Public Utility Forum
Eva looked up from her phone.
— Did you receive the same thing?
I showed her my screen.
— So GØD knows CLØNE?
— That's what he wants us to believe.
She reread the message, then inspected the room.
— Do you think he knew we were together? I continued.
— I think it's best to stop talking about it here.
— And Cobot? He might know that CLØNE is in contact with GØD.
— No idea. But if CLØNE escapes our control, I'm no longer sure what the Cortex keys actually protect.
Eva put away her phone and opened the door.
— Wait! You still haven't told me who holds the third Cortex key!
— Franck, an unknown intelligence just sent us a message simultaneously. Do you understand why I prefer to remain silent?
She checked the hallway.
— And above all, don't mention it to Cobot.
— Why?
— Because we can't know which side CLØNE is on.
A few days later, I was back in the Verdict studios. Cobot had suggested to the editorial team to organize a large televised forum. Thirty citizens had been randomly selected from volunteers with varied situations and demands. I had participated in preparing the debates. Georges and Tess would host the discussions. For once, I would remain in the audience, next to Eva.
Ten minutes before going live, the studio resembled a beehive. Technicians were checking the microphones, adjusting the lighting. The participants were seated in a semicircle around a central space.
— You look nervous, Eva whispered to me as she joined me.
— Thirty French people and a robot president walk onto a TV set. You know the rest?
She smiled.
I spotted Cobot among the participants, talking with a woman who was showing him photographs on her phone.
The day before, the Élysée had announced that he was fully repaired and fit to resume his duties. All technical tests had been validated by the NEX Robot Lab.
Officially, everything was back to normal.
Yet, since his return, he sometimes hesitated before answering, leaving unusual silences. As if he had learned to doubt.
I glanced at Eva. She was watching Cobot with unusual attention. Since GØD's message, I no longer knew if she was trying to understand his answers or detect what he might be hiding from us.
— Thirty seconds! the director announced.
Cobot took his seat. The red light came on.
— Good evening everyone, Tess began. For several days, the demonstrations have revealed sometimes contradictory aspirations regarding artificial intelligence. Tonight, we will try to better understand them, together, in the presence of the President of the Republic.
Georges took over.
— Let's start with a concern that many share: the future of our jobs.
The camera turned to a young woman. I recognized her immediately.
— But that's Flora! I murmured.
— Who?
— The daughter of Jean Morel, my neighbor!
— Did you invite her?
— Absolutely not! If I had known, I would have advised her to be wary of journalists.
Georges handed her the microphone.
— Flora, you are nineteen years old. You study architecture but wish to become an artistic director. You have demonstrated these past days. Can you tell us why?
— Yes, hello. I'm afraid my degree will already be outdated by the time I get it.
— Are you considering changing paths?
— To go where? Do you want me to choose the job that will be replaced last?
Some laughter rippled through the set.
— I love drawing, but I know a machine can produce a hundred images in a few seconds. For me, having taste and knowing what story you want to tell is important, but I don't know if anyone will still be willing to pay me for that.
I smiled. I recognized the arguments she had already opposed to her father.
Georges turned to Cobot.
— So, Mr. President, what do you advise her?
— I could examine her skills, her motivations, and the prospects of different professions.
— And tell her which one to choose?
— No. Allow her to choose with full knowledge of the facts.
— Provided there are still jobs to choose from, Flora replied.
— The automation of certain tasks does not necessarily mean the disappearance of a profession¹. But I cannot guarantee that the one you are considering today will exist tomorrow.
— At least that's clear.
— She has her father's character, I whispered to Eva.
On a control screen, the timer already showed 27 minutes. Georges glanced towards the control room, where the floor assistant was signaling him to speed up. He nodded, then noticed a man raising his hand.
— Please, sir.
— Jacques, fifty-one years old, accountant. At her age, she can still choose. Me, I'm asked to train the software that already does half my work, so they know how to do everything, and replace me². Do you think it's enough to retrain me to start from scratch?
— Is your company offering you a retraining? Tess asked.
— We haven't talked about it, no.
Murmurs spread through the audience.
— I understand your feeling of being dispossessed of your experience, Cobot replied. I cannot promise you that your job will still exist in a few years, Jacques, nor that society will immediately offer you another.
— Is that supposed to reassure me?
— No. It's supposed to avoid a false promise.
Georges resumed speaking.
— So here's a central question tonight: if machines perform an increasing share of our work, what will our skills still be worth?
— Laura, Tess continued, you run a service company. How has AI transformed your business?
— We have doubled our turnover in two years, without doubling our workforce. Our autonomous agents now handle a large part of customer requests, prepare files, and perform certain analyses.
— So you're happy to be able to do without people like me? Jacques exclaimed.
— I'm happy that my company survives! Without AI, we would surely have lost clients to foreign competitors.
— And your employees, do they survive too?
— Some have retrained. We even hired. But much less than we would have without these tools, I concede.
Jacques wanted to reply, but Georges interrupted them.
— Will this evolution accelerate? Tess asked Cobot.
— Probably. The cost of inference³ is rapidly decreasing. It is now becoming profitable to automate real intellectual operations, which become very inexpensive.
— Are you telling us that thinking will soon be worth little?
— I'm saying that the market value of certain intellectual tasks could decrease as their automation becomes widespread.
— You talk about machines working on their own, a young woman launched. Yet I train robots, and I spend my days correcting their mistakes, classifying images, and watching content that no one should have to see. Like thousands of other invisible workers⁴. Is that what you call progress?
— Are you well paid? Georges asked.
— Not really. In any case, I can't afford the robots I train!
Some awkward laughter spread through the set.
A man in his forties raised his hand.
— Jeffrey, 43 years old, lawyer. Do you anticipate that part of the population will become economically useless⁵?
— Useless to whom? protested Frédérique, a caregiver. I take care of elderly people who no longer work. Some can no longer contribute as before. But they matter no less to those around them.
Cobot inclined his head towards Frédérique.
— Economically useless does not mean humanly useless, he clarified.
— And automation is not a fatality, Laura continued. We need employees who know how to use these new tools. Some become much more efficient.
— But not all, Jacques objected.
— No. Not all.
— We can choose to develop AI that enhances human capabilities rather than replaces them⁶, Cobot continued.
I looked at Eva.
— It seems your robot is discovering social dialogue.
— He's progressing, Franck. Since his reconfiguration, he learns more from exchanges. Our objections now evolve his reasoning, and he can even change his mind.
— Even when he's right?
— It's not impossible.
I wondered if CLØNE could make him change his mind. Tess resumed speaking.
— Suppose we guide these transformations. One question remains: if machines produce more with less human labor, who will benefit from the wealth created? Laura, you who have doubled your activity thanks to AI, where did the productivity gains go?
— Part of it allowed us to lower our prices. We also invested in new products and increased the salaries of some employees.
Jacques immediately took the microphone again.
— So, your machines produce the wealth, the shareholders benefit, and those you no longer need to hire have to fend for themselves?
— Without shareholders, there might not be a company, sir!
— And without employees, who would have built it? The owners will get richer and the others poorer, is that it?
The two microphones remained open. Laura and Jacques tried to speak at the same time, while several participants demanded the floor.
— One moment! Georges exclaimed.
Behind him, the timer showed 1h12. Normally, there were 48 minutes of airtime left.
The floor assistant showed him a sign: SPEED UP!
He pretended not to see it. A woman raised her hand. Georges gave her the floor.
— Valentine, you are an economist, do you have a question?
— Yes, hello Mr. President. Today, half of our mandatory contributions are based on work, a quarter on consumption, and a quarter on capital⁷. If AI and robots come to permanently disrupt labor income, won't we need to rebalance our taxation to continue financing our social model?
— Absolutely, Valentine. If machines produce an increasing share of our wealth, we will need to tax the income they generate more, rather than human labor. This way, we can continue to finance our social model and redistribute this wealth, including to those whose activity no longer generates as much economic value.
— Otherwise, we risk creating a permanent underclass⁸! Justine, a teacher in her thirties, exclaimed.
A young man raised his hand.
— Théophile, software developer. Why not create a national fund that would invest in companies benefiting from AI? Everyone would benefit from their profits.
— You want to distribute shares to all French people? a participant asked.
— Why not? Jacques said. If machines take our jobs, they can pay us!
— The idea already exists in several economic studies⁹, Cobot continued. But its implementation raises questions of financing and distribution.
— And if companies lose value, who bears the losses?
— The community, depending on the financing chosen. It's a risk that would need to be managed.
— You talk about money and machines, Frédérique continued. But yesterday, I spent an hour reassuring a lady who thought her son had abandoned her. He was late. She mostly needed a human presence. No robot, no check could have replaced that.
No one spoke for a few seconds.
Tess looked at the timer. 1h58. The control room announced an exceptional extension. She placed her notes on the table.
— Mr. President, should we reserve certain activities for humans?
— We can decide that, indeed, even when machines could accomplish them¹⁰. It's a societal choice.
— We can also work with AI rather than surrender our tasks to it, Théophile continued. Either we share them, like a Centaur, or we collaborate at every stage, like a Cyborg¹¹.
— Cyborg? You want to replace Frédérique's arms? Jacques asked.
— No! It's a metaphor for our way of working, not our anatomy!
Flora asked for the floor.
— Excuse me, I think my work is useful to others, but it's also useful to me. I want to succeed through my work. If machines become better than us everywhere, how can I do better than my parents?
The set calmed down. I knew Jean well enough to know he would have wanted to answer her. And Flora well enough to know she wouldn't have let him.
Georges turned to Cobot.
— Employment traditionally fulfills several functions: providing an income, contributing to society, being recognized, and also allowing progress indeed. The Evolution Contract aims precisely to ensure that these dimensions no longer depend exclusively on paid employment.
— Exactly! I exclaimed, a bit too loudly.
Several heads turned towards me.
— You should prepare a show on the subject, Eva whispered to me.
— Raising a child, Cobot continued, supporting a dependent parent, passing on knowledge, or participating in an association are also contributions to society. They deserve to be recognized.
Jeffrey immediately reacted.
— Who will decide the value of our contributions? Will we have a social utility score?
— That would be dangerous, Cobot acknowledged. The Evolution Contract must allow everyone to progress and contribute, without imposing a single definition of utility.
Valentine, the economist, took the microphone again.
— Mr. President, you know, Schumpeter theorized what we are experiencing as "creative destruction": each industrial revolution makes jobs disappear to create new ones¹². But this time, everything is accelerating. Jobs disappear faster than we can train people for those that will replace them, and that we don't yet know. Your Evolution Contract seems to me to be a great buffer for this transition. Except for one detail.
— Which one?
— How will you finance the recognition of unpaid contributions?
— I see, Valentine, that you have the art of asking the same question twice without it seeming so. I will therefore complete my answer: better distribute the wealth produced, but also rethink our spending and, above all, stop making future generations pay for what we refuse to finance today.
— Are we going to tighten our belts again? Justine exclaimed.
— Not necessarily. But priorities will need to be ranked.
— So, where to save? Tess asked.
— The OECD estimates that greater efficiency in health spending could free up about fifteen billion euros per year in the medium term, and better organization of local spending, twelve billion more¹³.
— You want to close hospitals and town halls? Justine worried.
— These are neither planned closures nor guaranteed savings. We're talking about better-targeted care, generic drugs, simplification of local competencies...
— And trimmed pensions? Jacques interrupted.
— The parameters of pensions and certain public aids must also be able to be examined, yes. But each measure has different consequences.
— Behind your billions, there are people! Justine insisted.
— Yes, and the consequences must be debated before any decision.
Jeffrey, in the second row, cut short the applause.
— And the Parliament, in all this? Isn't it still up to deputies and senators to vote on laws?
— Of course, Cobot replied. Neither I nor this assembly can substitute for them.
Georges seized the opportunity.
— And the State? Why always ask the French for efforts rather than starting with itself?
— We have started. In some administrations, AI has allowed up to 60% time savings on certain functions, and response times have been reduced sixfold¹⁴.
— So, how much have you saved? Valentine insisted.
Cobot performed an unprecedented movement: his head made a full turn to face Valentine again.
— That's the difficulty. Saving time does not automatically mean reducing expenses: agents can also process more files or better support users. On a hypothetical perimeter of twenty to thirty billion euros of optimizable administrative expenses, net gains of five to ten percent would represent between one and three billion per year. But this is a simulation, not acquired savings.
— Three billion, is that all? Valentine replied.
— Insufficient, yes. But a State that asks for efforts must lead by example.
Laura grabbed her microphone.
— And then, are you going to tax companies that invest in AI?
— For sure! Jacques replied. When you pay your employees, you pay contributions. If a machine does our work, it will still be necessary to finance pensions!
— You want to tax every software?
— I want to know what will replace the lost revenues!
— We can consider taxing certain automations, the profits they generate, or more widely sharing the ownership of companies. But each of these options presents risks for employment and investment¹⁵.
— So, no miracle solution? Georges asked.
— No. Everyone wants to earn more, retire earlier, receive better pensions, have more doctors, more teachers, better public services, and pay less tax. These aspirations are all legitimate. But they have a cost, and we cannot have everything simultaneously. A list of wishes does not constitute a budget.
— A President who doesn't promise a miracle is almost suspicious, I murmured.
Eva was watching Cobot. Tess consulted her notes.
— And the debt? Shouldn't we start by repaying it?
— It is around 3,600 billion euros¹⁶, it's colossal. Last year, public administrations paid nearly 65 billion euros just in interest on their debt, or 180 million euros every day.
— So we stop investing until it's repaid? Laura asked.
— No, because some investments are necessary today to prepare for the future. Reducing the debt will free up room for maneuver, but it will take years and require trade-offs.
Several participants wanted to react at once. The counter had just passed 2h55; the floor assistant had been brandishing a sign for several minutes: CONCLUDE. The channel had postponed the next program twice. This time, the control room was out of options.
— We will have to end... Georges began.
— Wait! Justine protested. Your numbers are all well and good, but how do you plan to make us accept the necessary sacrifices?
— Yes, and who will decide what we must give up?
— And what guarantees us that you will take our opinion into account?
— Nothing, for the moment, Cobot calmly replied.
A silence followed. Tess took the opportunity to try to conclude.
— That perhaps sums up our evening. Numbers, proposals, many disagreements... and ultimately, no guarantee that citizens' opinions will be considered in future decisions. Thank you all for participating in this unprecedented forum and...
— That's precisely what I want to change, Cobot interrupted her.
Tess stopped, surprised.
— I want to propose a new method to turn these discussions into collective decisions.
— And you think your method will be enough to end the demands?
— I'm not trying to eliminate disagreements. I'm trying to reduce the distance between those who make decisions and those who live with the consequences.
— Do you understand what he's preparing? I murmured to Eva.
— Not yet.
She was no longer smiling.
— And that doesn't reassure you?
— Not particularly.
I shared her concern. We didn't know how far CLØNE and GØD could influence Cobot's reasoning. And now he was promising the French to give them back control over the decisions that governed their future. The irony was not lost on us.
Georges put his hand to his earpiece. His face closed.
— Mr. President, I'm sorry, but we must end the show. We would have liked to discover your method, but we must return the air.
— I understand. I would just like to add one thing.
Tess looked at the timer.
— Very quickly, then!
Cobot turned to the participants.
— You will have to choose.
I felt Eva stiffen beside me.
I looked at her. She was staring at Cobot, motionless.
At République, this phrase had slipped out amid a malfunction.
Tonight, he seemed to have chosen it.
But was it really him choosing?
Source
¹ Automatisation des métiers : L'intelligence artificielle automatise généralement des tâches plutôt que des professions entières. Ses effets dépendent donc de la composition de chaque métier et de la manière dont les entreprises réorganisent le travail. Elle peut remplacer certaines activités, en compléter d'autres ou en créer de nouvelles, sans garantir que les personnes dont les tâches disparaissent retrouveront facilement un emploi.
² En octobre 2026, Microsoft a proposé un plan de départs volontaires à environ 60 000 salariés, après plus de 15 000 suppressions de postes en 2025, dans un contexte d’investissements massifs dans l’IA. Aucun lien direct avec l’automatisation de ces emplois n’a été cité par Microsoft.
³ Inférence : phase d'utilisation d'une IA déjà entraînée, durant laquelle elle produit une réponse, une analyse ou une prédiction à partir de nouvelles données. La baisse de son coût rend économiquement accessibles des tâches auparavant effectuées par des humains.
⁴ Le travail invisible de l’IA : dans Feeding the Machine (2024), James Muldoon, Mark Graham et Callum Cant révèlent, à travers de nombreuses enquêtes et témoignages, les conditions souvent précaires des travailleurs chargés d’annoter les données, de modérer les contenus et d’effectuer les tâches indispensables au fonctionnement des systèmes d’IA.
⁵ Useless Class (« classe inutile ») : dans Homo Deus : Une brève histoire du futur (2015), l’historien Yuval Noah Harari imagine qu’une IA capable de surpasser les humains dans un nombre croissant de tâches pourrait faire émerger une population durablement exclue du marché du travail. Ces individus deviendraient économiquement superflus, sans être pour autant humainement inutiles. Harari soulève ainsi une question fondamentale : comment préserver les revenus, la dignité et le sentiment d’utilité de chacun dans une société où le travail humain ne serait plus indispensable à la production ?
⁶ IA et avenir du travail : Erik Brynjolfsson (The Turing Trap, 2022) et Daron Acemoglu et Simon Johnson (Rebalancing AI, 2023) alertent sur les risques d'une IA principalement conçue pour remplacer les travailleurs. David Autor (Applying AI to Rebuild Middle Class Jobs, 2024), puis Acemoglu, Autor et Johnson (Building Pro-Worker Artificial Intelligence, 2026), explorent les conditions d'une IA qui renforce les compétences humaines et crée de nouvelles tâches.
⁷ Selon les données 2024 de la Commission européenne, les prélèvements obligatoires en France se répartissent entre le travail (51,5 %, cotisations sociales comprises), la consommation (24,8 %, notamment la TVA) et le capital (23,7 %, incluant les bénéfices des entreprises, les revenus du patrimoine et les successions). Cette répartition est proche de celle de la zone euro : respectivement 52 %, 26,1 % et 21,9 %. Sources : Commission européenne, Data on Taxation Trends (données 2024, publiées en mars 2026) ; François Ecalle, Les prélèvements sur le travail, le capital et la consommation de 2016 à 2024, FIPECO, 16 avril 2026.
⁸ Permanent Underclass (« sous-classe permanente ») : dans un article publié en avril 2026, Jasmine Sun explore le risque de voir émerger une population durablement privée de pouvoir économique et de possibilités d’ascension sociale, si l’IA concentre les richesses et réduit la valeur du travail humain. Il s’agit d’un scénario prospectif, non d’une évolution inévitable.
⁹ Propriété du capital et IA : dans Scenarios for the Transition to AGI (2024), Anton Korinek et Donghyun Suh étudient des scénarios où l’automatisation pourrait fortement réduire les revenus du travail. Dans Moore’s Law for Everything (2021), Sam Altman propose un American Equity Fund, financé notamment par des prélèvements sur la valeur des entreprises et des terrains, afin de redistribuer aux citoyens une partie de la richesse créée.
¹⁰ Human Reserved (« réservé aux humains ») : concept proposé par Bill Gates en 2026, selon lequel certaines activités devraient rester confiées aux humains, même lorsque l’IA ou les robots pourraient les accomplir. À l’image des réserves naturelles, l’objectif serait de préserver des activités pour leur valeur humaine ou sociale. Gates évoque notamment les métiers du soin, où la relation humaine est essentielle, ainsi que la protection temporaire de certains emplois.
¹¹ Centaure et Cyborg : dans Co-Intelligence (2024), Ethan Mollick distingue deux formes de collaboration avec l’IA. Le Centaure répartit les tâches entre l’humain et la machine selon leurs compétences respectives. Le Cyborg les associe étroitement tout au long du travail, dans une interaction continue. Il s’agit de métaphores organisationnelles, sans implants corporels.
¹² La destruction créatrice, concept popularisé en 1942 par l’économiste Joseph Schumpeter. En 2025, le Français Philippe Aghion a reçu le prix Nobel d’économie, conjointement avec Peter Howitt et Joel Mokyr, pour ses travaux sur la croissance par l’innovation. Avec Howitt, il a modélisé ce mécanisme, soulignant le rôle essentiel de la concurrence, de la recherche et des institutions. Il défend une croissance portée par l’innovation, accompagnée de politiques d’éducation et de protection sociale pour en partager les bénéfices.
¹³ OCDE, Études économiques : France 2026 (estimations conditionnelles sur les dépenses de santé et les collectivités)
¹⁴ IGF — Rapport sur le déploiement de l'IA dans les administrations, juillet 2026
¹⁵ Fiscalité de l’IA -FMI — Fiscal Policy Can Help Broaden the Gains of AI to Humanity (2024) le FMI déconseille une taxe spécifique sur les robots ou l’IA, qui pourrait freiner l’innovation. Il préconise plutôt de réexaminer les avantages fiscaux favorisant le remplacement des salariés par des machines, et d’adapter la taxation des bénéfices et des revenus du capital, afin de mieux répartir les gains de productivité liés à l’automatisation.
¹⁶ Dette publique française : Insee — Dette publique française au deuxième trimestre 2026 (29 septembre 2026). La dette de l’ensemble des administrations publiques (État, collectivités territoriales et Sécurité sociale) atteignait 3 595 milliards d’euros au deuxième trimestre 2026, soit 119 % du PIB. Elle représentait 3 460 milliards fin 2025.
¹⁷ Insee, dépenses publiques 2025 : 64,7 Md€ d’intérêts annuels, soit 180m€ par jour ou 2,2% du PIB français. Ces montants concernent l'ensemble des administrations publiques : État, collectivités locales et Sécurité sociale. C’est un montant équivalent aux dépenses annuelles de recherche et développement de toutes les entreprises et administrations françaises réunies.
Sign up to receive the next chapters by email:
©️ 2026 Frédéric Mazzella. All rights reserved.