Ink has always belonged to the living. We spilled it from quids of steel, pressed it through lead type, and painted it across screens with the nervous energy of our own heartbeats. But a few months ago, the editorial board of Le Monde did something startling. They did not commission a column from a seasoned foreign correspondent, nor did they interview a tech titan in a glass-walled boardroom in Silicon Valley. Instead, they opened their prestigious pages to two dead ideologies speaking through ghosts.
They published the political rhetoric of Bernie Sanders and the populist battle cry of Steve Bannon. For a different look, see: this related article.
Not because either man had written a fresh manifesto for the Parisian daily. Rather, the newspaper fed their historical speeches, their television broadcasts, and their decade-spanning arguments into the silicon belly of a large language model. The instruction given to the machine was simple, cold, and profoundly unsettling: Synthesize these voices. Show us how you see us.
Read that sentence again. Similar reporting regarding this has been published by Ars Technica.
When you ask a statistical engine to channel a democratic socialist and a nationalist strategist, you are not engaging in journalism. You are running a seance. You are summoning the ideological extremes of the modern world and asking an unfeeling architecture of weights and biases to decide how they sound when they agree on the threat of autonomous machines.
To understand why a French newspaper felt compelled to stage this digital theater, you have to remember what it feels like to watch your own vocabulary get stolen.
Imagine sitting at your desk at midnight, staring at a blank document while an automated assistant in the sidebar cheerfully offers to finish your sentence. It gets the grammar right. It mimics your cadence. It even uses your favorite rhetorical trick—that dry, self-deprecating aside you thought was uniquely yours. For a fraction of a second, a cold shiver runs down your spine. Who wrote that? You, or the machine that learned how to be you by reading everything you ever published?
This is the invisible crisis facing modern discourse. We are no longer just building tools that write. We are building mirrors that distort our political and cultural lineage until every radical idea, every cry for labor reform, and every nationalist warning is flattened into polite, corporate syntax.
Le Monde recognized this vulnerability before it consumed them entirely. By publishing the machine-generated mashup of Sanders and Bannon, they were throwing a flare into the dark. They were saying: Look. Look at what happens when our deepest political convictions are processed through the meat grinder of stochastic parroting.
Consider the paradox. Bernie Sanders spends a lifetime railing against the billionaire class, arguing that the consolidation of capital in the hands of a few tech monopolies threatens the very survival of the working class. Steve Bannon preaches an apocalyptic gospel of national sovereignty, weaponizing cultural grievance against globalist elites. In the real world, these two men would not share a stage, let alone a cup of coffee. Their worldviews are tectonic plates grinding against one another, creating the earthquakes that define our elections.
Yet, inside the neural network, those differences dissolve.
The machine does not hate capital. It does not love the working class. It recognizes only probabilities. It looks at Sanders’ syntax—inequality, oligarchy, justice—and Bannon’s cadence—sovereignty, betrayal, elites—and blends them into a smooth, digestible slurry of algorithmic concern. The jagged edges of human rage are sanded down. The blood is wiped off the historical documents. What emerges is a ghost that sounds hauntingly authoritative, completely devoid of pulse.
This is what regulatory frameworks are trying to cage right now. When policymakers in Brussels, Washington, and Paris sit down to draft guardrails for artificial intelligence, they are not just arguing about copyright law or data privacy. They are fighting over the ownership of human expression.
If a machine can read Sanders and Bannon and produce a speech that feels more coherent, more rhythmic, and more persuasive than a human politician on a bad day, what happens to truth? What happens to authenticity?
Let us walk down a hypothetical street in central Lyon. It is raining—that fine, gray mist that makes the stone facades look like wet slate. In a small apartment above a bakery, a historian named Claire is staring at her laptop screen. She has spent forty years archiving political tracts from the French resistance, the student revolts of sixty-eight, and the bitter labor strikes of the nineties. She knows the smell of damp paper and mimeograph ink. She knows the sound of a crowd chanting in the street until their throats bleed.
She clicks a link to the Le Monde experiment. She reads the synthesized dialogue.
Then she closes her eyes and shivers.
"It's too clean," she murmurs to the empty room.
That is the verdict. It is too clean. Real political passion is messy. It stumbles. It contradicts itself. It is fueled by hunger, anger, and the desperate hope that tomorrow might look different than yesterday. An algorithm has never gone hungry. An algorithm has never stood in a freezing rain line outside a factory gates waiting to hear if its livelihood has been outsourced to a server farm in Ohio.
When we let machines frame our political debates, we sterilize the struggle. We turn class warfare and cultural anxiety into content generation.
The European Union’s push for AI regulation—the sprawling, contentious framework known as the Artificial Intelligence Act—is an attempt to erect a sea wall against this tide. It demands transparency. It requires companies to disclose when training data includes copyrighted works. It tries to force the ghosts back into the machine.
But regulations are written by lawyers using the language of yesterday to govern the technology of tomorrow. They are paperwork trying to catch lightning.
The real battle is happening deeper down, in the psychology of the reader. We are becoming addicted to synthetic eloquence. We want our opinions pre-packaged, our outrage curated, and our political fears articulated with the crisp, flawless cadence of a machine that never needs to catch its breath.
Step back for a moment and look at the broader picture. For centuries, literacy was the great equalizer. To write was to exist; to be published was to matter. Today, literacy has been outsourced. The average citizen no longer needs to struggle to find the right word to express their grievances against corporate greed or cultural displacement. They can prompt an app to do it for them in three seconds.
In doing so, we are renting out our souls to a server farm.
When Bernie Sanders talks about the exploitation of labor, he is talking about sweat, bone, and time—things that cannot be recovered once spent. When Steve Bannon talks about the erosion of national identity, he is channeling a visceral, ancient fear of displacement. These are human experiences, born of mortality and limitation.
An artificial intelligence has no mortality. It does not age. It does not fear death, and therefore it cannot truly understand what makes life worth fighting for.
When Le Monde published those ghost speeches, they were not offering a glimpse into the future of political discourse. They were holding up a mirror to our intellectual laziness. They were asking us if we are willing to let algorithms do our rebelling for us.
Because if we let machines write our manifestos, it will not be long before they start choosing our battlefields.
The rain continues to drum against the window in Lyon. Claire wipes her glasses, opens a new document, and begins to type by hand. Each keystroke is slow. Each sentence requires effort. She makes a typo, deletes it, and tries again. It is inefficient. It is exhausting.
It is wonderfully, terrifyingly human.
