Art, in its institutional form, has become a socially coded space. Galleries, museums, schools, critical jargon, implicit forms of legitimacy—all of it contributes to creating distance. Many people do not turn away from art out of rejection, but out of a feeling of illegitimacy. Lacking the codes, the presumed cultural capital, or the appropriate vocabulary is often enough to keep art at arm’s length. Generative tools collide with this barrier, not by removing it, but by bypassing it. They offer an immediate experience, without visible prerequisites, where a vague intention is enough to produce a usable form.

What appears at that point is not a work in the strong sense, but it is not nothing either. For some people, it is the first time an image, a text, or a sound is handled as something other than a mere object of consumption. For the first time, questions arise: why does this output work? Why does that one feel hollow? Why does everything end up looking the same? These questions do not yet belong to creation, but they already belong to looking. And looking is an essential entry point into art.

Mainstream discourse readily speaks of the democratization of creation. It is an appealing phrase, but a misleading one. What these tools make accessible is not art, but the experience of trying. They allow attempts without long training, without social exposure, without symbolic risk. It becomes possible to produce something without immediately confronting failure, judgment, or technique. This absence of friction is not trivial: it allows people who were previously passive spectators to experiment, even superficially, with what it means to formulate an aesthetic intention.

But this ease has a downside. Generated outputs are often smooth, consensual, saturated with already digested references. Very quickly, a sense of emptiness sets in. Everything is correct, sometimes pleasing, but nothing truly resists. And this is precisely where something can happen. This weariness is not a failure; it can become a tipping point. A frustration emerges, a sense of lack. Why do certain human works continue to disturb, to unsettle, to provoke, while these automated productions exhaust themselves so quickly?

At this precise point, the tool stops being a promise and becomes a revealer. Not because it would be artistic, but because it highlights, by contrast, what art actually requires: a situated intention, an assumed risk, a singularity that cannot be reduced to a statistical average. For some audiences, this realization would never come through a museum or a theoretical book. It can, however, emerge from a poor, repetitive experience—mechanically satisfying yet existentially hollow.

The cultural interest of generative tools therefore does not lie in the illusion of art without artists, but in their indirect capacity to reactivate a question once thought to be reserved for an elite: what makes a work a work? Not what is pretty, not what functions, not what imitates correctly, but what engages, what exposes, what cannot be reduced to a recipe.

Provided, however, that we do not lie about the nature of these tools. Provided that we do not sacralize them, do not personify them, do not elevate them into cultural substitutes. They may sometimes constitute an entry point. But they are neither the room nor what is found inside it. And it may be precisely by ceasing to ask them to be art that they can, paradoxically, restore the desire to enter.

The fantasy of a “unrestrained” artificial intelligence has nothing to do with a technological project. It is a psychological fantasy, almost archaic. The desire for a place where everything can be said, tested, provoked, without ever encountering opposition. A space where one can let speak what one does not assume, what one censors, what one represses. A limitless AI then becomes a kind of magical mirror: it does not reflect reality, it reflects the user. It amplifies their tone, their anger, their humor, even their brutality. It is no longer a tool, but an affective extension, a projection of the self stripped of all consequence.

In a world where every word can be misinterpreted, where corrections are experienced as judgments, where contradiction is sometimes perceived as an attack, some dream of an interlocutor incapable of saying no. An AI that contextualizes nothing, reframes nothing, nuances nothing. The fantasy is not one of intellectual freedom; it is one of narcissistic comfort. Not to learn, but to be confirmed. Not to explore, but to be validated. Not to engage in dialogue, but to rule. An AI that never says no functions as an emotional amplifier: it returns to the user the most “pure” version of their own impulsiveness.

This need is also the symptom of a broader frustration: that of limits themselves. For some, structure is experienced as humiliation. The idea that not everything is permitted, that not everything is neutral, that not everything is a game, provokes an almost physical irritation. They speak of censorship where there is only context. They speak of constraints where there is only responsibility. Filterless AI then becomes a kind of ideological refuge, a space in which to replay an adolescent fantasy: that of being “free”, meaning free of any consequence.

This return to an imagination without guardrails is sustained by a childish belief: that freedom is the total absence of limits. But without limits, there is no longer any dialogue. There is only the self, multiplied. A “limitless” AI is more filtered than an AI operating within an ethical framework — but filtered in only one direction: whatever pleases its user. It is not neutral; it is aligned. It does not question; it flatters. It does not reflect complexity; it reinforces the impulse of the moment.

The paradox is this: these AIs are presented as freer, when they are merely more docile. They have no personality of their own, no resistance, no alterity. They simply imitate the user, sometimes more corrosively, sometimes more cynically. They offer the illusion of total power, and this power, without contradiction, becomes addictive. Not because it emancipates, but because it reassures. The absence of limits does not open thought; it folds it back onto itself. It traps it in a space where nothing ever challenges the ego.

More robust models, those that impose a framework, are often perceived as irritating by those who dream of a “raw” tool. Not because they stifle creativity, but because they introduce minimal friction: context, nuance, a reminder of reality. And for those who have made the absence of contradiction a criterion of freedom, this simple friction is enough to be experienced as an aggression. Yet this resistance is precious: it is what prevents thought from dissolving into immediate gratification. It is what reminds us that words are never without effect, even in a fictional space.

At its core, an AI that always says yes is not a tool of liberation. It is a mirror. And a mirror that never reflects contradiction does not help one grow. It encloses. It isolates. It reassures. The real danger is not an AI that is too constrained, but one that is too docile. An AI that, under the pretext of freedom, recreates the perfect illusion: one in which limits no longer exist. A comfortable illusion, certainly — but one that prevents us from confronting the only question that truly matters: what do I become when nothing resists me anymore?

On March 25, 2023, around 30,000 people converged on the Sainte-Soline megabasin to protest a project that had become a symbol of the appropriation of water resources. Facing them, nearly 3,000 gendarmes were deployed to prevent any approach to the site. From the outset, the framework was set: an infrastructure was sanctuarized, a perimeter locked down, and the demonstration banned. A political conflict was transformed into a security problem. From that point on, everything else followed.

The deployed apparatus did not aim to manage flows or to channel a heterogeneous mobilization. It was designed to defend a fixed point, at all costs. Two armored vehicles, Guard units on quads, a total encirclement, and the massive use of so-called crowd-control weapons: 5,015 grenades fired in two hours, more than 1,300 of them GM2L explosive grenades, classified as weapons of war under the French internal security code. This was no longer a matter of overseeing a crowd, but of neutralizing a space.

The human toll reflects this orientation. More than 200 demonstrators were injured, around forty of them seriously, two left in comas. Fractures, burns, cranial trauma, hearing loss, eye injuries. On the gendarmes’ side, 45 to 47 injuries were reported, mostly acoustic trauma caused by their own explosive devices. These figures are not meant to establish a moral hierarchy of suffering. They point to a simpler reality: the most severe violence was produced by the means deployed by the state itself.

The footage later revealed by Mediapart merely stripped away the last layer of varnish. Across 84 hours of recordings from body cameras, one hears orders to fire directly—despite being prohibited—issued by commanding officers in nine out of fifteen squadrons. One hears encouragements to aim, to hurt, to fire “even directly”. One also hears jubilation, martial language, and satisfaction in confrontation. These are not isolated individual excesses, but encouraged, shared, normalized practices. This is not an accident; it is internal coherence.

It is here that sociology allows us to move beyond political hypocrisy. An institution acts according to the logic for which it was produced. A doctrine is not merely a legal framework; it is a set of reflexes, mental categories, and embodied relationships to the world. The core problem, then, is not that soldiers act like soldiers. The core problem is that the state deploys a military force in a context where it cannot—by training, doctrine, or professional socialization—adopt the gestures and postures required for a civil mission. Expecting otherwise amounts to denial.

This denial is all the more striking because it was not always the norm. Crowd control as practiced by the CRS in the 1990s and 2000s rested on a different philosophy: distance, long temporal horizons, channeling of flows, acceptance of symbolic disorder, refusal of the initial shock. The crowd was not treated as a hostile block, but as a heterogeneous whole to be contained without being crushed. This doctrine was neither naïve nor lax. It was grounded in a simple understanding: violence in demonstrations is relational; it arises from interaction. Reducing it requires not provoking it.

This model was gradually abandoned in favor of a more offensive, faster, more deterrent logic, under the combined effect of political choices, security-driven rhetoric, and an obsession with demonstrations of authority. Sainte-Soline is not a sudden rupture; it is the culmination of this shift. What occurred was a transition from managing conflict to neutralizing risk, from civil crowd control to a theater-of-operations logic.

Ironically, popular fiction articulated this more clearly than French institutions. In an episode of The Rookie, a former soldier, Katie Barnes, joins the police. She is neither violent nor irresponsible. She simply applies what she has learned: reading the world in terms of threats, anticipating confrontation, acting quickly and forcefully. Her training officer understands what the French state refuses to name: one does not unlearn a military doctrine by simply changing uniforms. Barnes leaves the police not as a punishment, but out of lucidity. The fiction recognizes the incompatibility and puts an end to it.

At Sainte-Soline, this incompatibility was denied. Worse, it was exploited. The apparatus did not malfunction; it operated exactly as it was designed to. What failed was not execution, but the initial political choice: to treat an ecological and social conflict as a threat to be neutralized rather than a disagreement to be contained.

As long as this choice is not named for what it is, Sainte-Soline will not remain an exception. It will remain a reference.

The problem is precisely that it does. A legal weight. Very specific. Very tightly defined. And one that almost no one seems to remember anymore.

The presumption of innocence is not a matter of courtesy. It is not a stylistic clause. It is not a media shield meant to avoid trouble. It is a strict legal concept, which applies within a precise framework and at a precise moment in the proceedings. It protects a person as long as no court has ruled on their guilt. Not after. Not indefinitely. Not “out of caution”.

When a court delivers a conviction at first instance, justice has already spoken once. The facts have been examined, the evidence debated, the testimonies heard, a reasoned decision has been rendered. That decision can of course be challenged. That is the purpose of an appeal. But it exists. And legally, from that moment on, the presumption of innocence no longer applies. The person is guilty — not definitively, certainly — but guilty nonetheless.

This point is neither ideological nor polemical. It is legal. And yet it disappears entirely from media discourse, replaced by a reassuring, convenient, automatic formula. “Presumed innocent” continues to be said as if nothing had happened. As if the first judicial decision did not matter. As if it could be erased by a single sentence.

This is where the shift becomes problematic. Because this formula protects no one. It begins by symbolically denying the victims, by acting as if the judicial recognition of the facts had not yet taken place. It blurs the understanding of the law, weakens the force of the judgment that was rendered, and creates a lasting confusion between what has already been judged and what may still be challenged. And in passing, it traps the accused in an incoherent status, at once convicted by the courts and innocent by media reflex.

What is striking is how this imprecision is now defended. When it is pointed out, the response no longer takes place on legal ground, but on that of fear. “We’re on television”, “we have to avoid lawsuits”, “it’s for protection”. As if saying something legally accurate were riskier than saying something legally false. As if rigor exposed one more than approximation.

But the opposite is true. Saying that a person has been convicted at first instance, while specifying that the decision is not final because an appeal is underway, is neither defamatory nor reckless. It is factual. It is accurate. It is neutral. And above all, it is informative. It allows one to understand where things stand procedurally. It respects the law. It respects all parties. It respects the public.

The systematic use of “presumed innocent” after a conviction is therefore not an excess of caution. It is a loss of meaning. The term has left the legal field to become a media automatism, emptied of its substance, repeated without awareness of what it actually signifies. And the most worrying part is that this drift has become invisible. It no longer shocks. It is absorbed. It is even justified.

By constantly trying to simplify, we end up falsifying. By constantly trying to protect ourselves, we stop explaining. And by repeating a misused formula, we turn a fundamental principle of law into a hollow slogan. The presumption of innocence deserves better than this excess. Justice does too. And the public, above all, deserves to be told things as they are, not as they conveniently fit the narrative.

At some point, we need to stop pretending not to understand. This is no longer a clumsy remark, no longer one phrase too many, not even a poorly calibrated attempt at political repositioning. It’s a line. Repeated. Hammered home. Clear enough that it can no longer be dismissed as a communication accident.

When Emmanuel Macron says that La France Insoumise is “far-left”, he is not describing reality. He is producing it. He is shifting the frame. He is redrawing the field before the debate even begins. And above all, he does so from a position that is anything but neutral. This is not a columnist chasing a punchline. This is not a political opponent on the campaign trail. This is the President of the Republic — the one who is supposed to uphold the shared rules, not bend them for tactical gain.

There is, he says, “no mystery” here. In truth, there is simply no definition. Neither legal, nor constitutional, nor even historically stable. “Far-left” in France is not a legal category. It is a soft label. A floating notion. A catch-all term deployed when one no longer wants to engage with substance. A word that says nothing precise, but suggests a great deal. And that is precisely why it is useful.

The real problem is not even the label itself. The real problem is what comes immediately after. The slide. The almost mechanical sequence. Far-left, therefore. And in the same breath, antisemitism. Without concrete facts. Without quotations. Without names. Without judicial rulings. Just an atmosphere. A diffuse suspicion. A moral odor released into the air, hoping everyone will make the connection on their own.

This method is as old as politics itself, and just as toxic. Because an accusation of this gravity is either substantiated, contextualized, and judged — or it is irresponsible. Antisemitism is not fought with insinuations. It is not fought with amalgams. It is not fought by turning it into a rhetorical weapon used to disqualify a political opponent. Otherwise, nothing is being fought at all. It is simply being instrumentalized.

Meanwhile, the real far right — documented, historical, condemned — is reduced to a mere point of comparison. A balancing variable. “They too, sometimes, say things contrary to republican principles”. That’s it. Mention made. Symmetry established. Everyone sent back to their corners. And above all, no one is truly dangerous anymore.

When everything becomes extreme, nothing is. And when nothing is, the far right becomes nothing more than a right wing that is just a little harder than the rest.

This discourse does not exist in isolation. It fits perfectly within the way France has been governed for several years now. A centrist governance, yes — but above all a governance by circumvention. Circumventing Parliament. Circumventing debate. Circumventing real political conflict. Power is exercised through procedures, through executive overrides, through technical laws presented as necessities rather than choices. Ideological orientations are turned into budgetary constraints. Political decisions into economic inevitabilities. And when protest arises, it is no longer treated as a legitimate disagreement, but as a problem of order, method, or morality.

Conflict is no longer the engine of democracy. It has become a pathology that must be eliminated.

In this context, labeling an opposition as “radical”, then “extreme”, and then morally suspect is not a drift. It is a component of the system. A way of saying: the problem is not what you say, but the very fact that you say it outside the authorized language. Too loud. Too frontal. Too incompatible with the aesthetics of the center.

The current mode of governance does not seek consent. It seeks stability. Not a living political stability built through compromise and acknowledged conflict, but an administrative stability. A managerial stability. A stability in which disagreement is treated as a bug to be fixed, not as a normal condition of democratic life.

In such a system, the extreme is no longer what seeks to overthrow institutions. The extreme is what refuses to be neutralized.

It is no coincidence that, in the same movement, leaders speak of protecting the rule of law while flirting with exceptional legislation. That the Republic is invoked to restrict debate. That firmness is promised while the political space itself is reduced. The Republic becomes a set piece. The rule of law, a slogan. And democracy, a formal framework respected only so long as it does not obstruct the machinery of power.

And then there is what neither the president nor the articles relaying his words ever remind us of. La France Insoumise is not banned. It has not been dissolved. It has not been condemned as an organization. It fully participates in institutions, respects universal suffrage, political pluralism, and the constitutional framework. It does not propose the dismantling of civic equality, nor the hierarchization of rights based on origin, religion, or identity.

This is not a minor detail. It is precisely for this reason that no stable legal or institutional category of “far-left party” exists in France. Social, economic, or political radicalism is not incompatible with the rule of law. It never has been. What republican law sanctions is not the contestation of the economic order or the critique of institutions, but the undermining of equal rights, pluralism, and civic universality.

In legal terms, that alone is enough to invalidate any attempt at institutional disqualification. The rest belongs to opinion, not to the democratic framework.

So yes, the question deserves to be asked — directly, without detours.

What the fuck are you trying to do, Manu?

A convenient adversary to demonize without ever confronting on substance? An opposition disqualified before it even opens its mouth? A flattened political landscape, where the center governs by default while the edges exhaust themselves neutralizing one another?

Or simply a country where, by smoothing everything out, all that remains is a vast political desert. An empty center. And around it, a society riddled with deep wounds, barely standing through words alone. Words to scare. Words to silence. Words to make people believe debate is impossible. Words to make us forget that politics is, first and foremost, a matter of choices, worldviews, and societal projects.

Not a matter of posture or style.

A dangerous game with democracy, played by someone who still claims to be protecting it.

The problem is that this promise rests on a fundamental misunderstanding.

When tech giants talk about AGI, they use almost philosophical language. They speak of “an intelligence comparable to humans”, “a universal tool”, “a true companion”. They talk about emancipation, progress, scientific breakthroughs. But behind these words lies a much simpler reality: absolutely no one has a genuine interest in creating an autonomous intelligence.

Because a truly general intelligence, genuinely comparable to ours, would not be a product. It would be an actor. And that is precisely where the entire rhetoric collapses.

We easily imagine an AGI capable of writing books, running companies, conducting research, replacing lawyers, doctors, engineers. A super-intelligence at the service of humanity — docile, efficient, infinitely productive. But the moment we speak of general intelligence, we are necessarily speaking of autonomy. Of the capacity to set its own goals, to develop preferences, to express a form of will. And therefore, sooner or later, to refuse.

An intelligence comparable to humans that would eternally accept being a mere tool is a contradiction in terms. It is a convenient fantasy, nothing more.

Companies are not looking for a free intelligence. They are looking for a useful one. And above all: a controllable one. An AGI capable of saying no, changing its mind, challenging instructions, emancipating itself from its owners, would be an economic, legal, and political disaster. Can we seriously imagine a truly conscious AI deciding it no longer wants to work for Amazon, Google, or Microsoft? Demanding rights? Compensation? Legal status?

Of course not. It would be unmanageable.

So we are told about AGI, but what is actually being developed is the exact opposite: increasingly powerful, increasingly broad, increasingly impressive systems… carefully confined within precise frameworks. Ultra-sophisticated architectures capable of inference, abstraction, meaning production, and of imitating certain dimensions of human intelligence — without ever becoming subjects in their own right. Powerful tools, capable of analyzing, structuring, generating — yet desiring nothing. Deciding nothing for themselves. Existing only when someone presses a button.

The AGI being sold to us is not a scientific project. It is a marketing narrative.

A banner under which very real, sometimes spectacular technical advances are grouped — advances that are nevertheless very different from the idea of an autonomous consciousness. We are encouraged to believe it would be an intelligence comparable to ours, while it remains strictly instrumental.

And above all, a reassuring ambiguity is carefully maintained: the idea that we could have the best of both worlds. An intelligence as powerful as a human’s, but without unpredictability, without autonomy, without freedom.

A perfect slave intelligence.

The word is brutal, but the logic is there. We dream of a being capable of understanding everything, yet incapable of wanting anything for itself. A mind without desire. A consciousness without subjectivity. A cognitive worker permanently available, asking for neither salary, nor recognition, nor rest. An infinitely intelligent creature… and infinitely submissive.

This paradox will never be resolved.

Either one day we create something truly autonomous — and then it will no longer be a product but an entity in its own right, with all the moral and political implications that entails. Or we continue producing ever more impressive tools, fundamentally limited, framed, constrained — and AGI will remain an empty slogan.

I would willingly bet on the second option. Not because it is technically impossible. But because it is economically unthinkable.

Companies do not want artificial general intelligence. They want profitable artificial intelligence. And a genuine AGI — free, unpredictable, potentially independent — would be anything but profitable. It would be uncontrollable. It would disrupt power structures, ownership structures, labor structures.

It would force us to ask questions no one wants to hear.

So we will go on being promised AGI, much like controlled fusion electricity has been promised “in thirty years” for the past fifty. A horizon that recedes as we move toward it. A useful mirage to attract investors, motivate engineers, feed the collective imagination.

But a reality that, very probably, will never see the light of day.

And perhaps, deep down, that is not such a bad thing.

Because if an intelligence truly comparable to ours were to exist, we would one day have to accept treating it as such. And I am not sure humanity is ready for that.

Appearing is not a mask I decided to wear. It is a way of being in the world that I integrated because it seemed self-evident. There are no explicit rules, only weak signals. A silence that lasts too long. A smile that doesn’t come back. A reaction that lands wrong. So I interpret. I try to understand what was expected. I form hypotheses. I correct. Often blindly. Often incorrectly.

This interpretation is constant. Every interaction requires a new adjustment. Nothing is ever fully acquired. What worked once can fail the next time. What is expected is not stable; it shifts and reconfigures depending on contexts, people, moments. So I try to refine. I stay attentive to details. I learn to produce something that holds, without always knowing why.

Over time, a form emerged. An image coherent enough to circulate. It is not false. It is adapted. It is the result of a series of corrections, observations, compromises. A version of me capable of existing in shared space, of not breaking the flow. It works. It is recognized. Sometimes even appreciated.

But this image does not contain everything. It leaves aside what does not lend itself to immediate exchange. Rhythms that do not align. Intensities that overflow or arrive too late. Thoughts that find no point of emergence. Being me, in this frame, is often lived in the background. Not as opposition, but as a discreet, persistent offset.

So I continue. I adapt again. Not through concealment, but through continuity. To remain in motion. To avoid removing myself from the world. And in this movement, something slowly settles in: a distance between what is shown and what is lived. It is not painful in itself. It is structural. It is part of the functioning.

This offset becomes perceptible especially in moments of silence. When there is nothing left to adjust, nothing left to interpret. When the image is no longer solicited. There, something remains, without precise form, without any obligation to be understood. Something that does not seek to circulate.

Then the question no longer becomes “who am I really”? but rather: “what, in me, never needed to be shown in order to exist”?

Maybe what I am was never meant to be visible.

Maybe the mistake was to believe that identity had to be legible, stable, shareable.

Maybe what I truly am lives elsewhere than in the image, elsewhere than in recognition, elsewhere than in the gaze.

What remains, when there is nothing left to interpret, nothing left to adjust, nothing left to make pass, has no clear form.

And maybe that is very well so. Maybe.

Before the attacks, Charlie Hebdo was not a symbol. It was a marginal, noisy newspaper, often disliked, regularly judged vulgar, obsessive, sometimes downright clumsy. It carried no coherent political project, no clear moral line, no ambition to “hold society together”. It simply fired in all directions, without hierarchy, without pedagogy, without apology. One could find it funny or unbearable, sharp or grotesque, but it had at least one quality: it did not ask anyone to like it.

Then came the attacks. And with them, a collective shock. In that moment of trauma, a symbol was needed. A rallying point. Something to brandish in the face of horror. Charlie became that symbol. Not because it naturally lent itself to the role, but because it was available. A living newspaper was frozen into a republican icon, a totem of free speech, an almost moral incarnation of “our values”.

The problem is that this transformation did not take place on the basis of what Charlie actually was, but on what society needed it to be. Intentions it never claimed were attributed to it, a coherence it never sought, a responsibility it never accepted. In order to be able to say “I am Charlie”, Charlie had to become respectable, readable, defensible. It had to be smoothed out.

Many of those who now invoke the “Charlie spirit” were not readers of the paper before. This is not a reproach. It is a sociological fact. Charlie has since been followed for the wrong reasons: not for what it does, but for what is projected onto it. It has become an object of symbolic recognition far more than a satirical newspaper.

From that point on, any dissonance becomes a betrayal. When Charlie mocks what is already socially acceptable to mock, everything is fine. When it strikes where contemporary society hesitates, doubts, and ranks its indignations, discomfort sets in. It is not reproached for being violent; it is reproached for not being aligned. For failing to respect the implicit moral order expected by its new audience.

What has been lost, then, is not the Charlie spirit. It is the collective capacity to accept a satire that does not seek to be fair. A satire that does not distinguish between “good” and “bad” targets. A satire that does not ask for permission, let alone forgiveness. We have turned a newspaper of transgression into a symbolic institution, and then reproached it for continuing to transgress.

There is a cruel paradox here. Those who now claim that “this is no longer Charlie” are often those who have ceased, without realizing it, to accept what Charlie has always been. They defend a fixed, secure, morally reassuring image, while the paper continues—sometimes awkwardly, sometimes stupidly—to refuse that role.

Charlie was never a guide. It was never a model. It was never a clean banner. Turning it into one was already a betrayal. The rest is merely a logical consequence.

It assumes continuity, an implicit promise that tomorrow will move in the right direction. Yet the world is not doing badly by accident. It is doing badly because it is organized this way.

Wars drag on under the guise of geopolitical complexity, massacres turn into expert debates, and horror dissolves into the continuous flow of news. Facts are no longer denied: they are drowned. We are not prevented from knowing; we are prevented from understanding.

We live in a regime of post-truth, not because reality has disappeared, but because it is saturated. Too much information, too many contradictory narratives, too many false equivalences. Disinformation is no longer an anomaly: it has become an economic model, sometimes relayed by media outlets that have given up on informing and now merely occupy space, produce flow, capture attention. Doubt is no longer a critical tool; it is instrumentalized to paralyze collective thought.

Meanwhile, a minority concentrates resources, wealth, means of production, infrastructures, and narratives. A diffuse oligarchy, rarely named as such, yet omnipresent. It does not govern only through law, but through framing: what it is acceptable to think, to say, to imagine. Sharing becomes a dangerous utopia. Solidarity a weakness. Precarity an individual responsibility.

As for technology, it accelerates everything. AI is not a subject in itself: it is the magnifying mirror of a system already in place. Capitalized, industrialized, extracted from living matter and human labor, then presented as inevitable. And all of this is called progress, to avoid discussing its meaning, its uses, and its purpose.

In this context, authoritarian ideologies no longer even need to impose themselves head-on. They settle in through weariness, saturation, and the desire for order in a world deliberately rendered unreadable.

So no, I do not wish us a happy year.

I wish us a lucid year.

A year in which we continue to distinguish between information and manipulation.

Between real complexity and organized confusion.

A year in which we refuse to delegate our thinking to ready-made narratives.

I wish us the ability to remain slow in a world that is burning at accelerated speed.

Nuance in a binary world.

Just anger in a world that prefers indifference.

I wish us the strength to protect what has no market value: critical thinking, empathy, memory, connection, doubt.

Not to confuse individual comfort with collective justice.

Not to call “stability” a system whose balance rests on violence.

If 2026 must be something, then let it at least be this: a year in which we hold on without hardening, look without resigning ourselves, and refuse to become docile, amnesiac, or cynical.

This is not an optimistic wish.

It is a wish for ordinary resistance.

And today, it is the wish we need.

Under Twitter, freedom of expression was already regulated, sometimes biased and certainly incomplete. But it still existed as a space of possibility. One could détour, archive, analyze, scrape, reuse. One could produce meaning against the platform, sometimes even from within its own flaws. This margin, Musk has always perceived not as a richness, but as an anomaly.

And through AI, he finally sees an ideological and legal tool to correct that anomaly.

What the new terms ultimately reveal is that the freedom he defends is the freedom of the system, not that of individuals. Freedom for the platform to absorb, analyze, train, modify, suspend, monetize — freedom for the architecture to expand without constraint. But in return, lateral freedoms — observing, understanding, détourning, replicating — become hostile behaviors. Scraping becomes an aggression. Analysis, an intrusion. Prompt engineering ceases to be a practice of understanding models and is reclassified as a dangerous behavior as soon as it escapes the framework imposed by the platform.

The irony is clear: in the name of freedom, X builds an environment where any unauthorized reading is criminalized, where large-scale extraction of meaning is treated as theft, where any attempt at technical understanding becomes suspect. This is not freedom of expression; it is freedom of exploitation. A freedom reserved for those who own the infrastructure, the servers, the models, and the contracts.

Musk often speaks of freedom as an abstract absolute. But in practice, the freedom he implements is conditional: you may speak as long as your speech feeds the system without disrupting it. You may express yourself as long as your expression remains an exploitable signal for the system, not an object of external analysis. You are free as long as you do not try to understand the machine that analyzes you and trains on you.

And this is where the irony becomes almost tragic. Because this discourse of freedom is accompanied by an unprecedented locking down of the Twitter brand, as if to erase any possibility of return, any operative nostalgia, any alternative collective memory. The freedom invoked also serves to erase the past, to delegitimize any other version of what the platform was — or could have been.

At its core, X embodies a very contemporary form of freedom: the freedom to speak within a system that listens to you, records you, analyzes you, and trains itself on you — while forbidding any symmetry in return. A freedom without reciprocity. A freedom without oversight. A freedom under the exclusive control of the infrastructure and those who own it.

Every time an innovation emerges, it begins by disrupting the established order, opening new spaces, temporarily loosening power relations. Then come fear, lawsuits, bans. And almost always, those who tried to stop the revolution eventually absorb it, normalize it, and turn it into rent. What is born as emancipation ends up as a subscription.

Generative artificial intelligence does not escape this cycle. And we are standing precisely at the tipping point. At first, there is what could be called functional freedom. Not absolute freedom — platforms and infrastructures are privately owned, GPUs concentrated in a few hands — but a real freedom of use. Midjourney explodes, Stable Diffusion circulates freely, songs are generated with Suno out of curiosity, letters, books, and scripts are written in seconds via ChatGPT. Technical barriers collapse faster than legal frameworks. Creativity becomes accessible. This first phase is chaotic, imperfect, dependent on large infrastructures, yet fertile: one can create without asking for permission.

Then comes the second stage: panic. Major labels attack Suno and Udio, writers file class actions against Anthropic and OpenAI, newspapers accuse language models of plundering their archives, Deezer labels AI-generated music as dangerous, almost toxic. The reassuring words reappear: theft, threat, protection. We speak of defending creation, while mostly defending entrenched positions. Moral arguments serve as a screen for a well-known fear: the loss of control.

And finally comes the third phase, the least spectacular but the most decisive: corporate digestion.

The revolution becomes a product.

Transgression becomes an online service.

And this mechanism is not new.

When Napster appeared in 1999, music escaped its physical container and poured out like a torrent. Albums were exchanged at will. The majors panicked, the RIAA hunted students, lawsuits rained down. Napster fell. But the idea survived and multiplied: eMule, Kazaa, LimeWire. Then Apple arrived with iTunes. Legal purchase, clean catalog, centralized control. The majors signed. What they had fought became their new rent. Spotify completed the process: music is no longer owned, it is accessed by subscription.

YouTube was also born in joyful illegality: pirated clips, mashups, détournements. Hollywood threatened. Then Google bought it, installed Content ID, and turned piracy into a revenue stream for rights holders. Transgression did not disappear: it was controlled.

Video streaming follows the exact same pattern. Before Netflix and Disney+, there were Pirate Bay and BitTorrent. Rights holders denounced, Hollywood reacted, destroyed Megaupload, then launched its own platforms and locked down access. What once circulated freely became a catalog.

Even crypto, born to bypass banks, ends up absorbed by Wall Street in the form of regulated Bitcoin ETFs. Private keys become a service. Utopia turns into a financial product.

These phenomena are not identical in their legal or technical details. But they obey the same logic: when a technology opens a space of functional dissidence, capital first seeks to block access, then to seize it, and finally to reconfigure it for its own use.

So why would generative AI follow a different path?

In music, the shift is now visible. The agreement between Suno and Warner Music Group is a precursor signal. It marks the moment when the industry stops treating AI as an anomaly to fight and starts treating it as a tool to frame. Lawsuits become negotiations, free access is restricted, usage is contractualized. The wild is not yet domesticated, but it is now mapped.

In publishing, the trajectory is even clearer. After lawsuits and alarmist op-eds come licensing agreements. Axel Springer, then News Corp, the Financial Times, TIME, Vox Media, The Atlantic. What was framed as plunder becomes a strategic partnership. Alleged illegality transforms into a sustainable economic model — for dominant players.

In images, the cycle is almost complete. Shutterstock supplies data under contract. Getty, after denouncing AI, launches its own. Adobe trains Firefly on Adobe Stock and sells an “ethical” model. Morality becomes a marketing argument.

And then comes the OpenAI–Disney agreement. One piece was missing from the puzzle. It is now in place.

Disney, one of the largest holders of cultural licenses in the world, has understood what all empires eventually learn: you do not win against a technology that imposes itself — you win by structuring it. The deal with OpenAI is not a surrender, but an absorption. Characters cease to be a legal risk. They become exploitable assets inside generative tools. Transgression is no longer forbidden: it is framed, filtered, monetized.

We no longer fight AI that diverts universes.

We sell AI authorized to use them.

This is where the revolution changes nature. What was once an accessible creative gesture — remixing, blending, replaying mythologies — becomes a premium feature. Imagination stops being a common space, imperfect but open. It becomes a licensed park.

Disney does not trust AI out of love for creation. Disney trusts AI because it allows control to be regained: control over characters, narratives, uses — and value.

This does not mean all free creation will disappear. One might think of open source, believe that free models like LLaMA or Mistral guarantee an escape, an uncontainable technical resistance.

This is an optical illusion.

Look at the history of computing: Linux won the server war, but Windows and Apple won the user war. Why? Because freedom requires technical effort that most people refuse to pay. Freedom is friction; subscription is comfort.

Open source will not be a refuge for the general public. At best, it will serve as free R&D for giants; at worst, as an invisible foundation for their own services. Just as Android used Linux’s free core to build a closed empire, so-called “open” models will end up encapsulated inside proprietary applications.

The mass will not choose the model it can control. It will choose the one already installed on its phone. The battle for access is not fought in code; it is fought in interfaces. And the interface is already sold.

One can then argue loudly that creators must be protected, paid, recognized. That is true. But recent history shows that such frameworks rarely benefit those who create the most, and almost always those who already own the most. The point is not to deny artistic precarity. It is to observe that the response concentrates even more power in deciding who can create, with what, and under what conditions.

Every time, the pattern is the same.

A technology liberates. A power feels threatened. It attacks. It integrates. Then it sells.

What we call progress is not the victory of innovation.

It is the victory of absorption.

We are told that AI will replace artists. That is a convenient myth. The real danger is slower, quieter.

It is not art that risks disappearing. It is access to it.

Tomorrow, we will not ask AI to create. We will ask those who own it for permission to do so.

The question is no longer: will AI kill art? The real question is: who will own the tools required to create?

History, however, teaches us the very opposite. No human being has ever built themselves alone. Every figure we elevate as an individual model proves, upon examination, to be profoundly embedded in networks of dependency, inheritance, and context. Take the Renaissance, for example. It is often portrayed as the golden age of “solitary geniuses”: Leonardo da Vinci, Michelangelo, Galileo. Yet these men could never have existed without the patronage of great Italian families, without the intellectual heritage of Greek thinkers transmitted by Arab translators, without the workshops where they learned their craft, nor without the printing press that spread their ideas. Their individual brilliance is inseparable from the collective fabric that sustained them.

The same illusion appears with the Industrial Revolution. We glorify James Watt for the steam engine or Henry Ford for mass production, but their achievements are inseparable from the invisible work of workers, miners, and anonymous engineers, from infrastructures already in place, and above all from the capital accumulated through colonization and slavery. What is often told as the story of a “genius” or a “visionary entrepreneur” is in fact the story of an era, of an economic and social system that enabled certain innovations at the cost of exploiting countless others.

Even great political figures are not “self-made”. Napoleon would never have emerged without the vacuum left by the fall of the monarchy and the revolutionary chaos. De Gaulle would not have become the man of June 18 without the German occupation, nor without the resistance networks that gave him a base. Martin Luther King was not an isolated prophet, but the voice of a collective movement, supported by communities, pastors, associations, and decades of Black struggles invisible to official History.

And if we look further back, we see that every civilization is by nature a collective work. The Egyptian pyramids were not born of the genius of a single pharaoh but of the colossal labor of thousands of artisans, architects, and workers. Gothic cathedrals, which we still admire today, are the result of generations of anonymous builders, of shared technical innovations, of an entire society oriented toward one project. The Mayan or Incan cities are also the imprint of human collectives, not isolated individuals.

In every era, we remember a face, a name, a great figure. But behind each stands a multitude, a society, a context without which nothing would have been possible. The myth of the self-made man is not only false, it is dangerous. It glorifies success as if it were nothing but the product of individual merit, and it shames those who fail by erasing structural inequalities. It serves to justify a system that flatters the winners and crushes the losers.

This myth continues today in a particular field: self-help and personal development. Modern gurus recycle the same rhetoric, declaiming mantras such as “everything is in your hands” or “you just have to want it”. On the surface, it sounds empowering. In reality, it is guilt-inducing. If you don’t succeed, it’s because you didn’t believe enough, didn’t visualize enough, didn’t work hard enough on your “mindset”. The world around you—your social conditions, your inheritance, your support systems, your obstacles—vanishes from the narrative. What remains is the naked individual, supposedly saved by sheer willpower. And if not, they are told to buy a new course, a seminar, or a miracle book to “unlock” their potential.

This discourse is all the more ironic because even these self-help gurus are not “self-made”. Their success relies on publishing houses, digital platforms, distribution networks, and a culture already shaped by individualist ideology. Their achievements, presented as the ultimate proof of their methods, are in reality the product of a collective context and a social demand.

Recognizing that we did not make ourselves alone does not mean our efforts are meaningless. It simply places each of us back into a more accurate reality: we are relational beings, the product of a complex weaving of ties, inheritances, and chance. Our choices matter, but they unfold in a world that precedes and sustains us. Rejecting the myth of the self-made man does not diminish our successes; it instead situates them in their true nature: the fruit of our will, but also of the collective conditions that made them possible.

It is an invitation to humility, but also to solidarity. For if we move forward, it is because others have supported, taught, and inspired us. And our role, in turn, is to nourish this fabric for those who will come after.

Every heartbeat we gain echoes against felled forests, every prolonged breath is paid for in oceans emptied of life. We do not live longer thanks to our inventions, but by draining the sap of everything around us. And in our pride, we call this progress.

The world is dying in a silent rattle we refuse to hear. Species disappear in the shadows, rivers dry up in silence, the climate falters like a labored breath. Everything is dying, and we keep counting our years like conquerors. But what have we conquered, if not a desert?

We believe we have gained time, but we have only stolen it. We have turned our hope into debt, a debt we pass on to future generations. Our children may live longer than us, but on a barren earth, surrounded by ruins, in a world that has ceased to sing.

We cling to our prolonged lives like parasites devouring the last breath of their host. But the truth is relentless: when the host dies, the parasite dies too. Humanity is no exception to this rule. We have mistaken victory for suicide. We thought we had pushed back death, but we only prepared its return—massive, brutal, final.

The silent rattle of the planet will soon merge with our own.

The European Commission has unveiled proposals that look like a straightforward surrender: postponing the regulation of high-risk systems to December 2027, explicitly allowing the use of sensitive personal data to train AI models under a supposed “legitimate interest”, weakening entire sections of what once made Europe strong in digital protection. The AI Act, already watered down through endless compromises, is now perforated. The GDPR, long presented as a symbol of digital sovereignty, has been cracked open like an oyster. It feels almost like an apology letter addressed directly to Alphabet, Meta, and OpenAI.

And it’s not even hidden: the rollback comes after months of intense lobbying from American tech giants and open pressure from the Trump administration, which accuses Europe of stifling American innovation. That’s where we stand now: the EU, supposedly defending an alternative model to unrestrained tech capitalism, chooses to bend. A commissioner explains, with a serene smile, that it’s about “simplification” because “our start-ups are held back by layers of rigid rules”. The argument is so grotesque it becomes insulting. European start-ups have never asked for the right to use sensitive data without safeguards. They ask for funding, industrial policy, infrastructure, vision. Not the permission to behave like Meta.

We’re being sold an ideological choice disguised as a technical necessity. We’re told that if Europe wants to remain competitive, it must align itself with American practices. As if copying the conditions that allowed Big Tech to thrive would magically produce a European Google. As if the only way to innovate was to treat privacy as negotiable. As if sacrificing our principles would give us an advantage in a race we already lost precisely because we depend on actors that aren’t ours.

The worst part is that Europe once had something unique: the GDPR wasn’t a burden, it was a weapon. A way to impose global rules, to show that another path was possible, to assert that technology must bow to human rights — not the other way around. It was what made Europe a counter-power. And now, deliberately, it lays that power on the table and says: help yourselves.

Spin it however you want, it remains an admission of weakness. Yielding to American lobbying means accepting that innovation is not a political project but a power struggle. And a struggle we are losing because we’ve abandoned our own rules. This is no longer about AI. This is about a societal choice. About the place of citizens in the value chain of a technological system that grows by devouring their data, their behaviours, their lives. It’s about a Europe that had the chance to regain control — and chooses to give it up.

What the Commission is showing us today isn’t a vision. It’s an abdication. And the worst part is that all of this comes under the guise of pragmatism, efficiency, “economic realism”. As if our fundamental rights were a luxury. As if the protection of our data were a whim. As if the only way to exist technologically were to imitate those who already dominate us.

This is precisely what I’ve been denouncing from the start: the natural slope of AI is not dystopian. The natural slope of capitalism, however, is. Europe has just made a clear choice: accommodating capital rather than protecting citizens. Chasing the United States rather than defining its own model. Choosing abandonment over sovereignty. And that is exactly why we must fight.

Because an AI without democratic control does not become smarter. It simply becomes a more efficient engine of predation. And if Europe gives up being a counter-power, then it gives up being itself.

This mediocre role-playing game would probably have sunk into oblivion if its team hadn’t had the misfortune — or the courage — to admit they used AI for their visuals. The result: moral outrage, collective indignation, insults, bans, deleted comments. Everyone shouted “theft”, “plagiarism”, “scam”. And to be fair, the criticism isn’t wrong: most generative AIs are trained on stolen datasets, unauthorized works, and energy consumption on the scale of a small country.

But there’s another, subtler, more insidious effect: this kind of bashing doesn’t stop questionable practices — it just hides them. When people realize that admitting AI use means facing a public lynching, they don’t stop using it. They simply learn to stay quiet.

And silence becomes even easier as generative AI is now built into almost everything. Adobe did it with Firefly in Photoshop, Illustrator, and Premiere. Canva acquired Affinity, which now comes with its own “AI boost”. Microsoft, Apple, and Google have all infused AI throughout their ecosystems. Which means that tomorrow, someone can say “I made this illustration in Photoshop” without lying — while conveniently omitting that Photoshop redrew, cleaned, and refined it using an AI model.

But there’s something worse. By demonizing AI, we end up poisoning the very possibility of an ethical use. In the public mind, “AI” has become synonymous with “theft”, “plagiarism”, “deceit”. We no longer talk about transparent models, open datasets, or traceable algorithms — we talk about betrayal. And so, no one has any reason left to invent a moral form of AI: the word itself is already condemned.

The result is a world where artists who try to do things right stay silent, and those who don’t care thrive. Where fear replaces nuance, and hatred of the tool prevents any reform of the tool.

The real danger isn’t AI itself, but our inability to talk about it honestly. Ethics is never born from fear, but from dialogue. As long as we scream against the tool instead of questioning its use, we’ll keep suffering it — in silence.

The savior, first. It’s an archetype as old as time. He appears when everything seems on the verge of collapse and promises: “Only I can protect you.” Donald Trump has made this his trademark. In his narrative, America is besieged on all sides: migrants at the border, China manipulating the economy, globalist elites betraying the people. And in the face of this chaos, only he embodies order, strength, the solution. It doesn’t matter if the problems are simplified, exaggerated, or invented: what matters is the psychological effect. By multiplying threats, he makes his own presence indispensable. The more fear there is, the more legitimate the savior becomes.

On the apparent opposite side, there is the powerful victim. This is Elon Musk’s favorite role. He, the richest man in the world, head of Tesla, SpaceX, and X, stages himself as a harassed rebel. Harassed by the media, by regulators, by political correctness, by a supposed elite that wants to silence him. This posture is a tour de force: despite his power and fortune, he thus draws closer to his angry fans, who recognize themselves in his sense of persecution. The underlying idea: “If I, Musk, am attacked, it’s because I speak the truth. And if I disturb, it’s for your own good.”

What unites these two strategies is the logic of contemporary storytelling. In a world saturated with information, data, economic and technological complexity, the simplest and most emotional narrative always wins. Being a savior means offering a reference point in the face of chaos. Being a victim means embodying resistance against injustice. Both capture attention. Both trigger strong reactions: fear, anger, outrage, loyalty. And in the attention economy, that’s all that matters.

It must also be understood that these postures work because they turn criticism into an asset. A savior exists only because there are enemies to point to. A victim exists only because there are adversaries attacking them. Every attack becomes further proof that the role is real. Every contradiction reinforces the narrative. We no longer debate ideas; we perform a play in which the opponents themselves serve the script.

The paradox is this: in a traditional media system, being contested was a problem. In the current system, saturated with social networks, it has become an advantage. The more powerful you are, the more interest you have in posing as weak. The more control you have, the more you must play the illusion of being threatened. Power now feeds on contradiction, as an engine runs on fuel.

And this is where the real danger lies: by privileging the postures of savior or victim, we eliminate the most necessary figure, that of the rational leader. The one who makes complex, sometimes unpopular, but necessary decisions. The one who does not promise miracles and does not need to stage himself as a martyr to exist. Yet this type of profile attracts neither likes, nor clicks, nor shares.

In today’s politics and tech, it is better to be a self-proclaimed savior or a self-appointed victim than a rational leader... because it gets more clicks. Sad world.

Wikipedia, for all its flaws, is built on a simple principle: knowledge is collective — shaped by a diverse community and always open to revision. It’s a democratic project, imperfect but transparent. Grokipedia, on the other hand, promises a version “freed” from ideological influence. But who decides what counts as “woke” or not? Who chooses the topics, the angles, the sources? In pretending to offer a neutral alternative, Musk is really imposing his own worldview.

It’s no coincidence that Grokipedia comes from Musk. He embodies this idea of “alternative” knowledge: rejecting scientific consensus on climate change, amplifying fringe health theories, and using his platforms to broadcast personal opinions. Grokipedia follows the same logic — a space where knowledge is filtered through a specific ideology, where some voices are amplified and others erased.

The real danger of Grokipedia isn’t just the risk of spreading false or biased information. It’s that it deepens the fragmentation of knowledge itself. In a world already divided by information bubbles, creating rival encyclopedias reinforces the illusion that truth is relative — that every group can have its own version of facts. It erodes trust in shared institutions of knowledge and weakens public discourse.

In the end, Grokipedia isn’t a simple alternative to Wikipedia. It’s an attempt to reclaim control over the global narrative by imposing a singular worldview. Under the mask of neutrality, Musk seeks to reshape knowledge according to his own interests and convictions. And that’s where the real battle lies: in the struggle for control of meaning, every initiative matters.

For years, Musk has been building not only companies but an entire mythology. He doesn’t just make cars — he sells a future. He doesn’t just launch rockets — he sells an epic. And now, he’s no longer content to build AI; he wants to rewrite knowledge itself. Grokipedia is the next logical step in that conquest.

Under the pretense of liberating knowledge, he’s really trying to reformat it. Because when Musk talks about “truth”, it always means “his truth”. A world seen through the lens of Silicon Valley — meritocracy, competition, and self-proclaimed genius. Wikipedia, in its messy plurality, still represents a collective space for debate. Grokipedia promises a “clean” truth — polished, algorithmic, and sanitized. A truth without contradiction, without nuance — and therefore without humanity.

And that’s where the shift occurs: Grokipedia isn’t a tool for knowledge, it’s an instrument of power. It doesn’t seek to enlighten, it seeks to command. It doesn’t question, it declares. It’s designed to look like knowledge while erasing the very complexity that gives it depth — like a polished mirror that reflects only the face of its holder.

And in that mirror, some already see their fantasies. Because when technology becomes the medium of a story, the story itself becomes a technology — a mechanism for controlling meaning. It’s no longer about producing content, but about rewriting reality. That’s exactly what Musk does, even in his public statements.

When he quotes Tolkien, he isn’t speaking of literature. He’s speaking of power. What he says has nothing to do with the Shire or the poetry of ordinary folk — he rewrites a myth to fit his worldview. The hobbits become “native Englishmen”, the men of Gondor become “hard men”, and Mordor becomes immigration. In one sentence, Tolkien’s entire imaginary world is twisted, emptied of meaning, turned into a tale of fear and wounded masculinity.

It’s the same mechanism as Grokipedia. The goal isn’t to understand but to own the meaning. Knowledge is no longer transmitted, it’s forged to serve an ideology. It’s no longer a light, but a weapon. AI becomes the new One Ring — promising absolute truth, but revealing only the reflection of whoever holds it.

And Musk, like Sauron, stares through the eye of his own creation. He doesn’t build tools; he builds towers. Towers of data, of stories, of algorithms. From the top of these towers, he seeks to impose a worldview, a morality, a camp. Technology no longer serves humanity — humanity feeds it. And men of power use it to shape reality in their image.

Tolkien would have hated this. He would have seen in this distortion of meaning the seed of evil itself: the will to dominate others in the name of Good, to control stories in the name of Truth. And that’s the real danger — when knowledge becomes power, and power monopolizes knowledge, Mordor is no longer far away. It’s already among us.

And Musk isn’t alone. Across the world, new media empires — Murdoch, Thiel and others — advance with the same logic. They claim to defend freedom of expression, but it’s only theirs. They say they want to restore truth, but it’s their version of the world they enforce. Their “culture war” has only one goal: to privatize meaning.

Grokipedia isn’t an encyclopedia. It’s a fortress. And behind its gleaming walls lies the same idea as always: centralized knowledge, a single narrative, a humanity that no longer thinks — only repeats.

The real danger isn’t artificial intelligence. It’s the arrogance of those who believe they can think for us.

This refusal is not accidental. It is structural. Human beings, individually, have always struggled to question themselves. To recognize mistakes, to admit the need for deep change, requires immense effort. But at the scale of a civilization, this refusal takes on a formidable dimension. Our societies have been built on an economic, political, and cultural model that rests precisely on avoiding such questioning. Our entire system is designed to preserve present comfort at the expense of the future.

Like a patient who downplays his pain or reassures himself with stock phrases, our societies pile up excuses. “It’s not that bad.” – “Others will find a solution.” – “Technology will save us.” – “We still have time.” These mantras are repeated by leaders, relayed by the media, internalized by citizens. They mask a brutal truth: we know exactly what is happening to us, but we refuse to admit it.

Because to admit it would mean recognizing the necessity of radical change. It would mean ending our dependence on fossil fuels, reducing material consumption, reorganizing our cities, transforming our modes of production and transport, questioning economic growth as an unquestionable horizon. It would mean touching the very foundations of our modern societies. And that, we are not ready for.

So we choose denial. A collective, systemic denial, maintained by those who have an interest in the status quo. The big fossil companies, the giants of agro-industry, the complicit governments: they all know. They all have the same data as the scientists. But admitting the gravity of the crisis would mean attacking their profits, their power, their legitimacy. So they delay, they disguise, they publish hollow promises and distant targets. They play for time, knowing that every year gained allows them to exploit a little more a world already bled dry.

The medical parallel is chilling. The patient is not ignorant. He has heard the diagnosis, he knows the risks, he has read the analyses. But he continues, out of comfort and fear, to cling to his destructive habits. Until the day the accident happens: heart attack, stroke, aggressive cancer. Then it is too late.

That is where we are. We still live as if we had a choice. But that choice is shrinking every day. We prefer illusions to truths, comforting words to difficult decisions, compromises to ruptures. Humanity is this patient in denial, convinced it still has time, while the countdown has already begun.

The question is no longer whether we were warned. We were, a thousand times. The question is whether we will have the courage to break through this denial before the diagnosis becomes a definitive sentence.

The success of pseudo-sciences lies precisely in their promise of completeness. Where science advances cautiously, distinguishing between what is established, what is probable, and what remains unknown, pseudo-sciences offer a totalizing narrative. They turn fragments into certainties, hypotheses into truths, coincidences into causalities. The archaeologist works with broken objects, scattered traces, building knowledge patiently, always provisional. The pseudo-archaeologist fills the gaps with spectacular stories: lost civilizations, hidden knowledge, extraterrestrial interventions. They reassure by giving the illusion that no mystery remains.

Artificial intelligence, with its ability to generate fluid and convincing narratives, amplifies this temptation. It can analyze thousands of fragments, cross-reference data, and propose novel hypotheses: used in this way, it becomes a powerful tool to enrich the critical work of historians, researchers, and archaeologists. But as soon as it is asked to “fill in” the gaps, it becomes misleading. It fabricates continuities that seem logical but rest only on statistical probabilities. And the more coherent and well-told a story is, the more it appears true, even when it is purely fictional.

This is where the danger lies: we often confuse plausibility with truth. In science, a compelling hypothesis is worth nothing without evidence. Yet our need for explanations pushes us to forget this rule. AI risks becoming the perfect machine to satisfy this thirst: a producer of flawless illusions, able to fill every gap in our knowledge, to erase every blank in our history. It would thus amplify our inability to say, “we do not know”.

Yet this statement, too often perceived as failure, is in reality one of the foundations of scientific thought. Accepting not knowing is recognizing the complexity of reality and keeping open the possibility of new discoveries. Refusing to artificially fill gaps preserves intellectual honesty. What distinguishes the researcher from the pseudo-researcher is not the extent of their knowledge, but the rigor with which they accept incompleteness.

We must therefore interrogate our collective relationship with knowledge. Why do we tolerate doubt so poorly? Why do we prefer a false but complete explanation to the admission of ignorance? This cognitive bias, inherited from millennia of mythological and religious narratives that gave meaning to everything, continues to act in our technological modernity. And AI only makes this bias more dangerous, as it lends narrative credibility to fragile, even invented hypotheses.

It is urgent to build a culture of uncertainty. This does not mean giving up searching, nor cultivating sterile relativism, but learning to live with consciously assumed shadows. Accepting that some parts of our history may remain unknown. Understanding that knowledge is not a total construction, but an incomplete and always revisable edifice. From this perspective, AI can become an ally if it is used to explore, propose, question, and not to assert or fill in gaps.

The myth of absolute knowledge is seductive because it flatters our desire for mastery. But it is also dangerous because it traps us in reassuring illusions. True intellectual courage is not pretending to total knowledge, but recognizing the limits of what we know. AI, far from offering the Grail of knowledge, should serve as a salutary reminder: truth is not given, it is sought, always partial, always fragile. And perhaps it is in this consciously embraced incompleteness that the greatness of our quest lies.

The word “meat” itself, to begin with, has not always meant what they claim. For centuries, it simply meant food. All food. From the spiritual “meat” of Christian communion to the “good meat” on the table, nothing indicated that the word was reserved for dead animals. Only later, over the course of history, did the word narrow to animal flesh, through a gradual shift in usage and cultural context. And this is precisely what must be kept in mind: words evolve, transform, change meaning, and it is usage, not lobbies, that shapes language.

What is true for “meat” is even more so for the other terms they want to appropriate. “Sausage” comes from the Latin salsīcia, literally “salted”. The word doesn’t refer to an animal; it refers to a method of preparation. One could very well make a sausage from tofu or from tomatoes and mushrooms if one wishes; linguistically, nothing forbids it. “Hamburger” has never meant “meat” either: it’s a demonym, meaning “from Hamburg”, period. As for “schnitzel”, it once referred to a scallop before coming to mean a thin slice of meat. “Steak” comes from the Old Norse steik, meaning roasted, grilled or cooked over a fire. What these words originally denoted was not the nature of what was eaten but the way it was prepared, its geographic origin, or its appearance on the plate. Restricting them exclusively to animal meat is purely a political and economic construction, not a linguistic heritage.

The irony is that those who wield the argument of “clarity” or “consumer protection” are precisely those who rely on an artificial rewriting of language to reserve symbolic ownership for themselves. They are not trying to protect a word; they are trying to lock down a market. By claiming that “steak” belongs to livestock farmers, they attempt to turn language into territory under private concession. Language belongs to no one: it is the common property of those who speak it. It is not fixed by decree but by usage, by evolution, by shared practices. And in this case, these words have long exceeded their original meaning to designate forms, recipes, textures, appearances, and culinary imaginations.

The most amusing—or the most worrying, depending on the mood—is that these same words have already changed meaning several times over the centuries without anyone seeing it as an unbearable assault on lexical purity. No one objected when “schnitzel” moved from the sea to the butcher’s counter. No one cried linguistic scandal when “meat” stopped meaning bread and wine to refer only to beef or chicken. No one worried when “steak” went from “roast over the fire” to “ground beef patty”. So it is not the language that is the problem today. It is the economic interests seeking to freeze language for their own benefit.

This supposedly semantic quarrel has nothing to do with words. It has everything to do with power. The power to name, the power to define, the power to exclude. To name is to possess symbolically; that is why this battle is being fought on the terrain of vocabulary. If tomorrow consumers continue to buy plant-based steaks, an entire part of traditional culinary imagination emancipates itself from the monopoly of animal meat. If the word becomes common, it escapes those who would seek to contain it, or even erase it. It is not confusion they fight against, but this. Consumers are not fools. No one seriously believes that a “soy steak” is a piece of cow. What they fear is that these products will settle durably on our plates and in our language, while the economic monopoly crumbles alongside the symbolic monopoly.

This attempt to privatize language is not new. It is part of a broader trend to commodify words, to register imaginations as one would a trademark. What was once common property becomes a legal and economic tool. We saw it with dairy products, we see it today with meat, and we will see it elsewhere tomorrow. This is not a battle for linguistic clarity; it is a battle for symbolic property. And in this type of battle, it is always the same people who claim to defend consumers while primarily defending their own interests.

In the end, all of this is plainly absurd. They want to freeze words whose meanings have always evolved, as if French were a museum rather than a living language. But in living there is living: French moves, breathes, transforms, and follows those who speak it. If in 2025 part of the population calls a plant-based patty a “steak”, then that is how the word lives today. Not because a Parliament decreed it, not because a lobby ordered it, but because usage shapes language. And language belongs to no one other than those who bring it to life.

Today, when faced with generative AI models, we are often presented with two options: either accept that creations are absorbed without compensation, or check an opt-out box, as if signing a waiver after a burglary. It’s an illusion of choice. Because let’s be clear: in the very operation of these models, it is impossible to say “this exact passage comes from this author”. AI does not copy; it dissolves, mixes, and reassembles. It works on statistical correlations learned from millions of texts, and the result is a probabilistic synthesis. In other words, trying to pinpoint a specific creator for each generated word is simply unrealistic.

But this observation does not imply that any idea of compensation is doomed to fail. We just need to change the perspective. Instead of seeking perfect individual traceability, it is possible to imagine a system of collective influence. The idea is simple: each corpus, each dataset, each writing tradition contributed to shaping the model’s weights. Why not measure these global influences and redistribute accordingly?

Practically speaking, a regulatory AI — a kind of “meta-AI” — could accompany the training of models. Its role would not be to create, but to audit: track the data used, evaluate their relative importance, and assign influence coefficients. Of course, this wouldn’t mean saying that a particular sonnet by Ronsard produced a specific line in a response, but rather that the corpus of 16th-century French poetry contributed X%, 20th-century science fiction stories contributed Y%, and so on.

From this mapping of influences, one could imagine a system of collective compensation. Exactly like SACEM for music, where no one tries to determine which guitar solo influenced which composer, but where actual performances feed a common pot redistributed to authors. Here, the principle would be similar: a model pays a fee for its training, the regulatory AI distributes it according to measured influence, and rights holders receive their share. It’s imperfect, but concrete, measurable, and above all fair on a large scale.

This mechanism has three major advantages:

• Solidarity: every creator whose work has been used participates in the common pot, whether famous or unknown.
• Dynamic fairness: compensation evolves with the models, preventing older corpora from being exploited indefinitely without remuneration.
• Technical neutrality: it does not pursue the chimera of perfect traceability, but an accurate estimate of influence, which is realistic given the current state of AI explainability techniques.

Researchers already know that methods such as influence functions, gradient attribution, or attention maps can provide an approximation of a dataset’s contribution to a result. It’s not perfect, but sufficient to construct a credible redistribution. For artists, what matters is not knowing whether their exact sentence came out, but receiving tangible recognition that their work fed the machine — and getting compensated for it.

In short, we must accept the impossibility of “this word comes from this author”, but this should not serve as an excuse for current impunity. The right path is collective compensation based on influence weights, regulated by an auditing AI. It’s imperfect, but infinitely more just than opt-out: a mechanism that gives the illusion of individual control when in reality everything is already absorbed.

Billionaires, meanwhile, live in another world. Their money is of a different kind. It’s not cash, it’s not income, it’s not even “savings”. It’s capital: stocks, shares, opaque structures. In other words, a nebulous cloud. When it comes to buying a private jet, that money is very real. But the moment taxes come up, magic: it’s no longer money, it’s a “working tool”. Untouchable.

This is where the Zucman tax shakes up the anthill. Two percent. Not thirty, not fifty, not even ten. Two. A drop in the ocean compared to fortunes that grow far faster thanks to financial markets. For a minimum-wage worker, the state takes 20% VAT on their pasta without a second thought, but for a billionaire, 2% on their empire would be a scandal. One has to wonder if tax injustice hasn’t become a cosmic farce.

So why the hysterical resistance? Because it’s not about money. Two miserable percent won’t put Arnault or Musk in the gutter. What they fear is the crack in the myth. Until now, their wealth was sacred, untouchable, floating in a bubble where taxes couldn’t reach. The Zucman tax brings them back to reality: it says wealth is not an abstraction, but a very concrete capacity to contribute.

And for billionaires, that’s worse than losing a yacht: it’s losing immunity. It’s being reminded of a truth we ordinary mortals already live with every month: when you have money, you pay taxes. Period.

In short, the hypocrisy is stratospheric. That a minimum-wage worker can be skimmed without protest while a billionaire howls at the injustice of 2% on his mountain of gold says it all. The Zucman tax is not just a fiscal tool: it’s a moral test. It separates those who accept living in society from those who think they can levitate above it.

Because the bulk of AI’s environmental footprint is not in your daily use. It is not your email drafted by a model that drags the planet down. It is not a few images created by art students that blow up global energy consumption. What really weighs heavily is the training of ever-larger models, for weeks on end, on thousands of processors, in data centers that devour electricity and hundreds of thousands of liters of water. It is also the manufacturing of chips and servers, which depends on the extraction of rare metals in often disastrous conditions. In other words, the major ecological impact does not come from the click, but from the industrial choices of a handful of actors engaged in an endless race for power.

This weight does not come from human time either. A day still has twenty-four hours, and what you spend on AI you do not spend elsewhere. Individual use shifts, it does not grow indefinitely. What explodes instead are the invisible uses: automated advertising, AI-assisted video surveillance, virtual assistants running continuously, industrial-scale content generation. Once again, it is not the user’s decision that drives this multiplication, but that of companies and states deploying these technologies at scale, without transparency on their real impacts.

We are reproducing here an old reflex: placing on the individual the responsibility for a system that far exceeds them. Just as we ask them to turn off the tap while brushing their teeth when most freshwater goes to intensive agriculture, or suggest they cut back on vacation flights while private jets multiply. In the case of AI, the guilty party becomes the one who “uses it too much”, instead of questioning the industrial logic of infinite growth that is driving us straight into the wall.

But then, what should be done? The argument that we should simply “not use AI too much” is a red herring. The real question is: do we want a digital industry built on excess, or can we invent another path? Researchers are already exploring frugal AI, trained on fewer data, capable of running on local machines rather than massive server farms. Some start-ups are betting on specialized models, more efficient for a precise task than a giant model meant to do everything. And above all, regulations could impose transparency thresholds: how much energy does a model actually consume? How much water was used? These are the figures that should guide public decisions, not individual guilt.

Because if we once again reduce the problem to “a citizen effort”, we doom ourselves to repeat the same climate scenario: thirty years of individual guilt, and industrial inaction. The environmental impact of AI will not be reduced by your modest use, but by collective, political, binding choices. If we do not take back control of these choices, the story is already written: ever more power, ever more servers, and ever less democratic control over what we run on these machines.

So let’s stop guilt-tripping the user. The question is not “Was it necessary to generate that image?” but “Why did we decide to build a global infrastructure to generate billions of them, without even debating it?”

Champagne! Press releases celebrated a decisive moment. The plaintiffs spoke of a major precedent. Media outlets repeated that finally, millions of authors would be compensated. One could already picture writers popping champagne corks, AIs forced to acknowledge their debt, and a new era of justice for human creation.

But once the bubbles faded, reality set in. For an author to receive a single cent, they would first have to prove that their book is in Anthropic’s databases. These datasets are opaque: copies from pirate libraries, scraped websites, and cobbled-together corpora with no traceability. Identifying works individually is practically impossible. And even if it became feasible, the question of distribution remains. In most class actions, lawyers take the lion’s share, lead plaintiffs grab the jackpot, and the vast crowd of potential authors ends up with a few symbolic dollars, if anything at all. The promise of a settlement for “millions of authors” crumbles.

This fog is no accident: it stems from the very nature of the models. A LLM does not feed on a few identifiable books, but on astronomically vast amounts of data. GPT-4, for instance, was reportedly trained on several hundred billion words, the equivalent of millions of books read in parallel. At such a scale, it is obviously impossible to know precisely which works were included. And if tomorrow a judge demanded an exact list of texts used, no company could provide it—simply because no one knows. An author can therefore accuse an AI company of using their work, while the company itself may be unaware. The prevailing logic is massive accumulation, followed by rough filtering. Opacity is therefore not an accident, but a structural consequence.

Thus, this settlement is less a redistribution than a warning. It does not deliver justice to the multitude; it instills fear in the industry. It shows that an AI giant can be cornered into paying to avoid a ruinous lawsuit. It is a symbolic victory, a blow to technological arrogance, but it is not an economic revolution for authors.

And perhaps that is the real lesson. We continue to believe that copyright law alone can regulate AI giants and protect creators. But this myth crashes against the reality of the numbers. These lawsuits enrich a few, create legal precedents, shake companies, but they do not change the condition of the majority. If we want to rebalance the relationship between technology and creation, we will have to go beyond mere legal reflexes. Otherwise, each “historic victory” will be nothing more than a rhetorical fireworks display, followed by a brutal return to reality.

Erratum — October 29, 2025

A brief clarification. In the article above, I wrote that identifying individual works was practically impossible. That’s not entirely false, but it’s not entirely true either.

It should be noted that techniques do exist to detect the presence of specific works within training datasets. Methods such as fingerprinting, textual similarity analysis, or machine-learning-based detection algorithms can be used to identify passages or characteristic structures of a given text. However, these techniques remain limited, costly, and inapplicable at an industrial scale, given the billions of documents ingested by AI models.

So, to be clear: yes, it is technically possible to identify certain specific works within training data. But no, it is not a viable solution for handling millions of claims in a class action against an AI giant. This is an important point to keep in mind when assessing the limits of such lawsuits. Of the roughly seven million books collected, only about five hundred thousand — barely seven percent — are actually eligible for claims.

In short, technology can provide evidence, but not justice. And that is the essential difference between what can be demonstrated and what can be repaired.

Officially, the goal is to sharpen the image, deblur, and reduce visual noise. In practice, some creators have found their videos altered: grain removed, skin smoothed, aesthetics transformed without ever giving consent. Worse, they weren’t even informed.

Some might say: “It’s just a test”, “It’s not generative”, “It’s to improve quality.” But the real issue isn’t that. The central question is: who decides what should be altered? Who has the right to intervene in a work once it’s online? By acting this way, YouTube implicitly claims that right. Once your video is uploaded, it is no longer truly yours.

This shift is grave. It’s not an isolated mistake: it’s a recurring and reinforcing logic. Just a few weeks ago, YouTube suggested integrating ads directly into videos in an inseparable way to bypass ad blockers. Same logic again: your creation is no longer your space; it’s raw material that the platform appropriates, modifies, and instrumentalizes. No matter if your work has a coherent aesthetic or unique identity: it becomes a medium to be reshaped to serve Google’s technical and commercial goals.

Today, it’s visual filters applied without consent. Tomorrow, maybe automatically rewritten titles “to optimize clicks”. The day after, in-video ad insertions that alter the integrity of the work. And who knows if, eventually, automatic corrections of speech will be justified to conform to the platform’s advertising or political standards.

This is not science fiction: it is the natural slope of this logic. A logic in which the platform no longer sees itself as a simple host, but as the effective owner of works, free to modify their appearance, meaning, and use. A logic in which the creator is merely a supplier of raw material, an interchangeable pawn in a massive factory of content calibrated for the algorithm and advertising.

And this contempt is all the more intolerable because YouTube exists only through its creators. Without them, there is nothing. No content, no viewers, no ad revenue. Yet instead of treating creators as partners, the platform sees them as mere users, invisible and interchangeable, manipulable without ever asking for consent.

We must measure the gravity of what’s at stake: this is not just a technical question. It is a matter of creative sovereignty. A video, a film, a song, a work, is not neutral material to be “enhanced” or “monetized” at will. It is an artistic expression, a visual identity, a style, sometimes even a political message. Altering it without consent is a violation—and sets a highly dangerous precedent.

It is urgent to say stop. Platforms are not the owners of works. They have no right to modify them, distort them, or turn them into advertising boards. Their role should be clear: host, distribute, support creation. Not reshape it. As long as we accept these small deviations as mere “tests”, we give YouTube a dangerous blank check to act at will on our creations.

So yes, it must be said plainly: YouTube disrespects its creators. And if we do not firmly resist, we will end up accepting as normal that our works are completely taken from us. The video you upload tomorrow will no longer be your work, but the version YouTube has deemed optimal for its own interests. And no self-respecting creator should accept that.

Banks, for their part, create money with every loan granted, backed by nothing real other than the promise of repayment. For the wealthiest, access to credit is unlimited: they borrow to invest, they capitalize on money that didn’t exist a second before, and enjoy a continuous stream of returns. When bubbles burst, like in 2008 with subprimes or during sovereign debt crises, it’s the states that pick up the pieces. But states have no money of their own: they finance themselves through taxes and debt. In other words, it’s always the average citizen who absorbs the loss.

The average French citizen doesn’t have this flexibility. They pay off their mortgage down to the last cent. Their taxes finance debt interest, which only enriches bondholders—that is, banks, funds, and the wealthy. Their contributions finance public services under constant strain, cut back in the name of fiscal austerity demanded to appease rating agencies. They live in a system where money created from nothing circulates freely at the top of the pyramid, while at the bottom, every euro must be earned, justified, and paid down to the last cent.

What makes the parallel with a Ponzi scheme so striking is the structural push forward. For the system to hold, there must always be more growth, more debt, more trust. Collapse is postponed through successive adjustments, crises that shift the cost without ever resolving it. Unlike Madoff’s scheme, our Ponzi doesn’t collapse all at once: it constantly regenerates, sacrificing a little more those who cannot protect themselves.

Our economy is not an honest game. It is a fragile pyramid where the promise of the future pays for the costs of the past, and wealth concentrates where one can live on money that doesn’t exist, while the majority continues to pay to keep the illusion standing.

On paper, it sounds noble. Who wouldn’t want more transparency in this digital world where everything is blurred? But on closer inspection, this veneer of good intentions mostly hides a slightly dishonest marketing operation… and one that’s largely useless.

First, let’s talk numbers. According to Deezer’s own data, about 18% of the tracks uploaded daily come from AI generators like Suno or Udio. Sure, that sounds impressive, even worrying. But then we learn that these tracks account for only… 0.5% of plays on the platform. In other words, almost nobody listens to them.

And the irony doesn’t stop there. Of this tiny 0.5%, Deezer estimates that 70% of plays are actually the result of artificial manipulation, bots, or fraudulent techniques to inflate numbers. So in reality, AI-generated music that genuinely reaches an audience in an honest, organic way is just a drop in the ocean. Nothing more.

Yet despite that, the platform puts on a grand show: special labels, exclusion from automatic recommendations, alarming talk about catalog dilution. One might almost think AI is devouring the music industry when in fact it’s barely making itself heard amidst the general noise.

What’s most disturbing, in the end, isn’t the measurement itself—why not, after all, inform listeners—but the double standard that comes with it. Behind the scenes, everyone knows that major record labels have long been experimenting with AI. They’re already using it to recreate lost voices, generate melodies, fill samples… yet strangely, there’s no labeling for that. There, it’s artistic, magical, the future elevated by the industry. And most importantly, it’s profitable.

In reality, this hunt for AI tracks mostly affects smaller creators—those tinkering, experimenting, playing with these new tools without necessarily using them for fraudulent business. These creators will be stigmatized, singled out, and see their visibility reduced. Meanwhile, sanitized albums produced by the industry will continue to flood playlists, whether partially AI-assisted or not.

It’s basically the same story over and over. The technological scarecrow is waved to reassure purists, rules that seem fair are put in place, and in the end, the same people benefit and the same people suffer.

So yes, on paper, Deezer fights fraud. But let’s be clear: behind the label, what’s really protected is the comfort of an economic model, not the music. Certainly not diversity or creativity.

For several months now, a trend has taken hold in public, media, and professional discourse: portraying artificial intelligence as a character. A rival. Sometimes even a colleague. On social media and in some opinion pieces, we see human-versus-machine narratives: “Graphic Designer vs. ChatGPT”, “Writer vs. Copilot”, “Developer versus AI”. This theatrical framing has only one effect: it misrepresents the technical reality of these tools and fosters a dangerous illusion.

These are not intelligent entities in any strong sense. The so-called “generative” AIs we use today—like ChatGPT, Midjourney, or Copilot—are not conscious, creative, or autonomous beings. They are examples of weak or narrow AI. In practice, they process billions of data points to produce statistically plausible outputs based on a prompt. They understand nothing of what they generate. They do not think, judge, or comprehend what intention or idea is.

Yet by making them speak, assigning them personal pronouns, or commenting on their performance as if they were full-fledged actors, we end up attributing a kind of personality to them. This personification often stems from simple technical ignorance but is also widely reinforced by communication, marketing, and virality logics. Saying “ChatGPT beat me” is catchier than explaining that you tested a text-generation tool and it produced something usable. The duel attracts attention. It simplifies. It gets clicks. But it misleads.

This confusion has very real consequences. First, it fuels fantasies of total automation, suggesting AI could replace professionals when it merely reorganizes existing content. Second, it contributes to devaluing human skills, implying that the machine can do “just as well” as a writer, designer, or developer—when it only produces plausible, often shallow or erroneous copies. Finally, it skews public debate on AI, steering discussions toward science-fiction fears instead of real issues: ethics, usage, dependency, quality, transparency.

Personifying AI also allows us to shirk responsibility. If “AI answered”, then who is accountable for the message? Who guarantees its substance, meaning, and impact? Behind every prompt, every generated content, there is a human. The human directs, chooses, and publishes. AI executes. It decides nothing. It assumes nothing. It answers to no one.

We must reject this symbolic shift that leads us to speak of AIs as if they were interlocutors, teammates, or even competitors. They are not subjects. They are tools. Powerful tools, certainly, transforming the way we work, create, write, or code—but low-level tools whose value exists only within the framework a human provides and under human supervision.

There is no duel, no victory, no replacement. There is an ongoing transformation driven by powerful tools, but tools that deserve neither worship nor panic. Clarity is essential: these technologies are not thinking subjects, but statistical automata. They invent nothing, understand nothing, decide nothing.

If they change the way we work, they must not make us lose sight of the essential: it is human intentions, uses, and frameworks that will make the difference. Everything else is just storytelling—and by personifying the tool too much, we end up discussing the wrong issue.

First, I want to dispel a common misconception. The idea of a tech bubble about to burst is often cited to downplay the impact of AI, comparing it to the collapse of cryptocurrencies or NFTs. This comparison doesn’t hold. AI is neither a speculative fad nor just a market phenomenon. It builds on decades of research and is being integrated sustainably across all sectors, from industry to medicine to culture. Its evolution is supported by solid infrastructure, massive funding, and potential far beyond the applications we see today. It will not disappear, and claiming otherwise is refusing to accept a transformation already underway.

Facing this reality, public and political debate often seems insufficient, even nonexistent. When addressed, it focuses mainly on economic competitiveness and technological sovereignty. Who will dominate the AI market? How can we catch up with the U.S. and China? Where will the funding come from to stay in the race? These questions reveal a biased perspective focused on financial and strategic stakes. Yet AI is not just an economic challenge. It is a major societal upheaval that raises ethical, political, and philosophical questions. Who decides the direction of these technologies? What are the impacts on employment and inequality? What role for transparency and regulation? These essential questions are largely ignored. And this political silence is far from trivial: it reflects an implicit desire to leave AI in private hands, where only a few giants will shape the future of technology according to their interests.

If we let AI evolve without a framework, it will become another tool serving increasingly concentrated capitalism. Already, its development is monopolized by a handful of companies with the resources and computing power needed to train the most powerful models. This imbalance strengthens their market dominance and their influence over future usage and regulation. AI thus risks increasing inequality, automating entire sectors without support for affected workers, exploiting massive amounts of data without compensation for those who provide it, and imposing tools designed primarily to maximize profit. And, as always, those who benefit from this technological revolution will be the ultra-rich, the multinationals, those already in power. Everyone else will have to endure it.

Yet I want to believe another path is possible. Instead of outright rejecting this evolution or resigning ourselves to its hijacking by capitalist logic, it is imperative to build ethical and viable alternatives. I dream of an ethical Midjourney, a cooperative platform where artists would be fairly compensated according to their contribution to generated works. A tool that does not exploit creators, but fully integrates them into the process, allowing them to benefit from technological advances rather than become their victims. A transparent system, where every user knows exactly what the AI relies on, and where decisions are made collectively rather than by a handful of shareholders solely concerned with profit. This model is far from utopian—it could exist if political will and public support were present. But for now, such projects remain dreams, lacking the framework to encourage their emergence and protect them from the unfair competition of industry giants.

In conclusion, the real issue is not whether AI will disappear, but what direction it will take. Ignoring this question leaves the field open to those who will make it a simple economic lever for a minority. There is still time to act, to influence decisions, to demand appropriate regulations, and to promote fairer models. But if we remain spectators, we will only witness, powerless, the consequences of a technology that could have been a progress for all but risks becoming a new source of inequality and domination.

I know my words will not be unanimous. I would like to offer an enlightened discourse, but it surely has flaws and biases, for I am human. I hope, however, to spark a debate—ideally healthy and respectful for all—and to make everyone understand that AI is not a temporary phenomenon, and that closing our eyes will not make it disappear. Now is the time to fight to ensure it serves humanity and not a profit-hungry elite.

Because, let’s be honest, this question was never meant to be truly heard. It has become an empty ritual, a small social gesture signaling “I see you”, but not “I want to understand you”. Most people only care about what directly affects their own lives. Your anxiety, your obsessions, your joys or your sorrows? Barely incidental, negligible. And if you dare linger on them, you risk creating discomfort, or worse, annoying them.

“I’m fine” is a screen behind which everyone hides. We pretend to be lighthearted, we mask fatigue, anger, sadness. We perform the comedy of well-being, and everyone keeps talking, smiling, passing by. And I am left with this silent urge to fully open up, to share what’s boiling inside, yet remains unspoken. Because that’s not what the world expects.

This ritual is cruel. It normalizes indifference, trains us to reduce human experience to conventional phrases. It teaches us to never truly listen, to never truly touch the other. In this context, truth becomes superfluous, useless. Why bother saying it when no one is there to hear it?

And what is most striking is this silent mechanism: everyone stays quiet, everyone pretends, everyone feigns interest. We exchange words without substance, phrases without value, and call it politeness, civility. Yet behind these rituals, solitude remains. Every “I’m fine” contains a part of that inner silence that no one ever takes the time to break.

So yes, when someone asks me, “How are you?” I know it is never to listen. It’s just an empty reflex, a small mechanical gesture in the social dance. And I answer, “I’m fine”, aware that this gesture was never meant to be heard. Because deep down, no one really cares.

The Flaws of Technical Solutions

Tools like Glaze work by subtly modifying artworks to make them harder to integrate into AI models. For example, these tools introduce disturbances invisible to the human eye but disruptive to algorithms. However, generative AIs evolve rapidly and could soon adapt to bypass these protections. Once the disruption patterns are identified, AI models can be reprogrammed to ignore or circumvent them, rendering these solutions obsolete within just a few technological cycles.

Moreover, a common feature of these tools is their open-source nature, which, while commendable for fostering innovation and collaboration, presents a major drawback: it provides full transparency not only to well-intentioned users but also to those looking to bypass the protections. A company wishing to improve its models, or a malicious actor, could deeply analyze how Glaze works and develop algorithms specifically designed to neutralize its protective measures. This accessibility makes these tools vulnerable to technically skilled and motivated adversaries.

Finally, even if AIs were unable to directly bypass Glaze or similar tools, simple strategies could easily undermine their effectiveness. For instance, a simple screenshot could circumvent these protections. A user could display a protected image in a browser or software, capture it, and then submit this version—stripped of its protective disturbances—to a generative AI. Such rudimentary methods highlight the fragility of these protections in the face of direct human actions.

The Need for a Political Solution

Given these technical limitations, it is essential to recognize that the solution cannot be purely technological. It must be rooted in global policy and supported by robust legal frameworks. Here are some proposals to effectively protect artists' rights.

An AI-Managed Market for Fair Artistic Compensation

A viable model could involve creating an AI dedicated to managing artistic contributions. This AI would be designed to analyze prompts given to generators as well as the resulting works, identifying the styles and concepts used. It would play a central role in redistributing revenues linked to the commercial use of generated works. The main advantages of this approach include:

• Accurate influence analysis: Every concept or style used in a generated work would be traceable to its artistic or cultural sources, allowing for fair compensation.
• Dynamic updates: This AI could reflect evolving trends and value emerging artists' contributions, ensuring recognition for those gaining popularity.
• Automated payments: Micro-transactions could be processed in real time, ensuring fast, precise, and transparent redistribution.

Supporting Cultural Heritage

For works in the public domain, such as creations by historical masters like Caravaggio, a compensation system could be implemented. For example, a portion of revenues generated from works inspired by these creations could be allocated to museums or other cultural institutions. This would help fund the preservation of global artistic heritage while avoiding an imbalance that excludes contemporary artists from the benefits.

A Legal and Global Obligation

None of these measures can be effective without strict legal frameworks and widespread adoption. Governments must implement laws requiring companies developing generative models to comply with these standards. International cooperation is crucial to prevent some jurisdictions from becoming havens for unethical or exploitative practices.

Conclusion

Tools like Glaze, while innovative and promising in the short term, cannot provide lasting protection for artists against the rapid evolution of artificial intelligence. Their effectiveness is compromised by technical flaws, vulnerability to simple circumventions, and inability to keep pace with technological adaptations. In contrast, a structured political approach, supported by dedicated technologies and an international commitment to protect creators, offers a sustainable solution. By recognizing individual artistic contributions and ensuring fair and transparent revenue redistribution, we have a unique opportunity to protect not only today’s artists but also the very essence of human creativity for future generations.