"The first act of resistance is to name things in order to reclaim power. What follows is the discipline of curiosity, the resistance to fatigue and indifference."
— Asma Mhalla, Interview with Pierre Coutelle, Librairie Mollat, 24/10/2025
March, D-35, five weeks before the election
Zara took a deep breath before speaking.
— I worked at Capgemini for six years. Data science. Projects for the Ministry of the Interior, Social Security, a few banks. We built predictive models. Nothing special, on paper. Algorithms to optimize flows, reduce fraud, improve administrative efficiency.
She paused.
— In 2023, I was given access to a classified project. The extension of the Fichier des titres électroniques sécurisés. Do you know what that is?
Gabriel shook his head.
— The largest database in France, said Malik without looking up from his keyboard. Fingerprints, identity photos, signatures. Every person who applies for a national ID card or a passport. Sixty million people.
— The extension, Zara continued, involved adding a behavioral layer. Analysis of micro-facial expressions at service counters. Cross-referencing with transport data, purchases, Wi-Fi connections. The idea was to build a risk profile for every citizen. Not based on what they do, but on what they might do.
She drew a circle on the whiteboard.
— The project was officially abandoned. Too controversial. But the models had already been trained. The infrastructure already built. They just changed the name and distributed it across several agencies.
Gabriel studied the faces around the table. Kenza, arms crossed. Malik, fingers still. Solène, pen raised.
— Are you certain? he asked.
— We've seen it, said Malik. The data. The servers. The connections. No need to guess when you can read.
He turned his screen toward Gabriel.
— Look.
The screen showed a map of France. Glowing points, intersecting lines, zones of color pulsing like a heartbeat.
— What is this?
— The Dileviathan, said Malik. Well, a small part of it. The data flows between ministries, telecom operators, banks, insurers, the major platforms. Each point is a server. Each line is a transfer of information. The color indicates the type of data.
Gabriel leaned in. The map was vast. Thousands of points. Millions of lines.
— This is... everyone?
— Not yet, said Malik. But it's coming.
He zoomed in on a region. Lyon. Red, blue, green points.
— Red is health. Blue is transport. Green is communications. Watch.
He clicked on a red point. A window opened. Lines of code, IP addresses, timestamps.
— That's the Croix-Rousse hospital. All patient records. Connected to the CPAM system. Connected to the pharmacy database. Connected to the medical transport registry. Every time someone enters that hospital, their identity, their diagnosis, their treatment, their insurance, their address, their phone number are transmitted to six different agencies.
— Without consent, Zara added.
— The law says otherwise, said Malik. But the law is written by people who don't understand technology.
He pushed back his chair.
— Come. We'll show you the mirror.
The corridor was narrow. Bare concrete walls. Cables running along the ceiling, fastened with cable ties. The air was saturated with dust and ozone.
Zara opened a door.
The room was small. Six screens arranged in an arc. Cables everywhere. A smell of warm plastic.
— Sit down, said Zara.
Gabriel slid into the seat. The screens were black.
— The mirror, said Zara, is a machine we built to visualize what the Dileviathan knows about you. Not in theory. In real time.
She pressed a key.
The screens came to life.
Gabriel first saw a shape. A network of glowing points, like a constellation. Then the points became words. Dates. Numbers.
— That's you, said Zara.
His name. Gabriel Moreau. Date of birth. Place of birth. Current address. Previous addresses. Social security number. Passport number. Carte vitale number. Bank account number. Phone number. Telecom operator. Phone model. Operating system version.
— These are the data found in public records, said Zara. What the State knows about you. But watch.
She clicked.
New points. New words.
List of incoming and outgoing calls. Duration. Frequency. Contacts. Call locations. Text messages. Content of text messages (excerpts). Emails. Senders. Recipients. Subject lines. Attachments.
— The operators, said Zara. They keep everything. The law mandates a one-year retention period for metadata. But operators keep the data longer. For resale, advertising, analysis.
The screen displayed his calls. His texts. His emails.
A message he had sent to his mother three years ago. I'm exhausted. I'll call you tomorrow.
— How...
— It's in a Data4Good file, said Zara. A company that buys data from operators for predictive marketing. Your message was anonymized, but not enough. We can reconstruct it.
She clicked again.
New points.
Purchases. All purchases. The bread from the corner bakery. Train tickets. Books on Amazon. Clothes from Decathlon. Medications from the pharmacy. Flowers from the florist.
— Bank cards, said Zara. Every transaction is recorded. Not just the amount. The location, the time, the type of business. With this, we can reconstruct your life. What you eat. What you read. What you wear. Who you see. Where you go.
The screen scrolled. Month after month. Year after year. His life.
The day he had bought a bouquet of flowers. February 14, 2022. At a florist near the place Bellecour.
He remembered. He had forgotten Valentine's Day. He had bought the flowers in a rush, between two meetings.
— You forgot Valentine's Day, said Zara. And you bought the flowers at 6:47 p.m. Just before the shop closed.
Gabriel turned to face her.
— You're reading my mind?
— No. We're reading your data. It's worse.
She clicked again.
New points.
Medical data. Blood tests. X-rays. Consultations with the GP. Consultations with specialists. Hospitalizations. Allergies. Current treatments. Test results.
— The hospital, said Zara. Your shared medical record. Connected to the CPAM. Connected to your mutuelle. Connected to your pharmacy.
The screen displayed his blood tests. His cholesterol levels. His blood pressure. His antibiotic allergies.
— This is... this is illegal.
— No, said Malik. It's the law. The DMP, the shared medical record, has been mandatory since 2022. You signed a consent form. But you signed without reading. Everyone does that.
Zara clicked.
New points.
Location data. GPS. Wi-Fi. Cell towers. Daily routes. Frequented locations. Time spent at each place.
— Your phone, said Zara. It sends its location every fifteen minutes. Even when you're not using an app. Even when you're at home. Even when you're asleep.
On the screen, his routes. His home. His office. The café he went to in the morning. The gym he went to in the evening. The supermarket where he did his shopping.
The journey he had taken to get there appeared. From Valence station to the hillside.
— You followed us, he said.
— No, said Zara. We found you. Through your data. We knew you would come. We just waited.
Gabriel stood up.
— This is a show of force, he said. You want to show me that you can know everything about me. That I'm transparent. That I have no secrets.
— No, said Zara. We want to show you that you have no secrets. Not from us. From them. The system. The architecture. The Dileviathan.
Zara echoed him.
— You have a smart shower, Gabriel. It measures the duration of your shower, the water temperature, the consumption. This data is sent to your energy provider, who resells it to a marketing company, who cross-references it with your health data to determine your stress level.
— What?
— You take longer showers when you're stressed. It's well known. Insurers know it. Some are beginning to use this data to adjust premiums.
Gabriel gripped the armrest and sank back down.
— You have a connected fridge, Zara continued. It sends your shopping list to your supermarket. Which resells it to food companies. Which uses it to offer you products tailored to your nutritional profile.
— You have a smartwatch. It measures your pulse, your sleep, your steps. This data is sent to your health insurer. Who uses it to calculate your "wellness score." The more you exercise, the lower your premium. But if you stop, it goes up.
— You have a connected car. It sends your location, your speed, your fuel consumption, your braking. This data is sent to your auto insurer. Who uses it to calculate your "driving score." If you brake too hard, your premium goes up.
Gabriel stared at the screen. The constellation of points. His life. His entire life.
— There is no conspiracy, said Zara. No plot. No secret meeting where men in suits decide to surveil you. It's worse than that.
— What?
— It's the architecture itself. The system is designed to collect data. Every device, every service, every application is a point of entry. The data accumulates. The algorithms analyze it. The models use it.
— And no one decides, said Solène. That's what's most frightening. No one needs to decide. The system decides on its own. It optimizes. It predicts. It steers.
— We call it steering, said Malik. Not manipulation. Not propaganda. Steering.
He came to stand in front of the screen.
— Imagine a drone, he said. Not a military drone. A delivery drone. It has a destination. It has to deliver a package. But it doesn't follow a fixed route. It adapts. It avoids obstacles. It takes the best paths.
— That's what the Dileviathan does with people. It has a destination. A direction. It wants you to think a certain way, to vote a certain way, to consume a certain way. But it doesn't force you. It steers you.
— How? asked Gabriel.
— By controlling your attention, said Solène. Recommendation algorithms. Social networks. Search engines. News sites. Everything is designed to capture your attention and direct it toward certain content.
— Not fake news, Malik clarified. Not disinformation. Orientation. It shows you what reinforces your beliefs. It hides what contradicts them. It creates bubbles. Echo chambers.
— And you believe you're thinking for yourself, said Solène. But you're thinking within the limits of your bubble. You can't see what lies outside.
The image of his parents came to him. His friends. His colleagues. Caught in the same net.
— And people? They don't realize?
— Some do, said Zara. But most don't. Because the system is invisible. You can't see the algorithms. You can't see the bubbles. You just see what you want to see.
— And by the time you start to see, said Malik, it's too late. Because the system has already modified your behavior. You've already changed. You've already become the person the system wanted you to be.
Gabriel stared at the screen. His life. His entire life.
The points. The lines. The zones of color.
His name. His date of birth. His purchases. His calls. His movements.
The shower he had taken in Paris, before he left everything behind. Twelve minutes. Thirty-eight degrees.
— You're right, he said. I no longer have a home.
Solène came closer. She touched Gabriel's shoulder.
— This isn't inevitable, she said. We can resist.
— How?
— We sabotage.
She moved toward another screen. A laptop, open, resting on a plastic crate.
— Not the infrastructure. We sabotage the model. We inject noise. We create contradictions in the training data. We make the algorithm doubt its own predictions.
She pointed to the screen.
— We call it the Parliament of Ghosts.
The screen showed an interface. Profiles. Photos. Names. Biographies. Hundreds. Thousands.
— Fake profiles, said Solène. Created by code we wrote. Each one has a story. A family. A job. Hobbies. Political opinions. Religious beliefs.
— Are they convincing?
— Very. We used AI-generated photos. Biographies we crafted ourselves. Simulated browsing histories. Consistent interactions.
— They talk to each other, said Malik. They comment on articles. They share posts. They form groups. They debate.
— And the algorithm watches them, said Solène. It learns from them. It builds profiles. It makes predictions. But the predictions are wrong, because the profiles are false.
— You're creating noise.
— Exactly. We sow doubt. We make the algorithm less certain of itself. We force it to make mistakes.
— And it works?
— A little, said Malik. Not enough. But we're improving.
The screen scrolled through the profiles. Faces. Names. Lives.
One profile surfaced. A man. Forty-five years old. Married. Two children. Engineer in Lyon. Practicing Catholic. Votes right.
Another. A woman. 32 years old. Single. No children. Lawyer in Paris. Atheist. Votes left.
Another. A student. 22 years old. Studying for a master's in philosophy. Lives with his parents. No religion. Votes green.
— Are they real? asked Gabriel.
— No, said Solène. They don't exist. But the algorithm can't tell the difference.
— And what do you do with them?
— We make them interact. We have them comment on political articles. Share posts. Debate.
— Why?
— To create a counter-model. A model that isn't aligned with the real world. A model that will learn false things.
— And then?
— Then we inject this counter-model into the system. Through public APIs. Through social networks. Through news sites.
— And the system becomes less reliable?
— Yes, said Malik. A little. Just enough for the predictions to be less precise. Just enough to give people a chance to think for themselves.
The thousands of profiles scrolled across the screen. The thousands of lives.
The campaign. The rallies. The speeches. The promises. All of it now floated behind the screen, like another layer of data. The system. The architecture. The Dileviathan.
— It's a guerrilla war, he said. Not a war.
— Yes, said Solène. It's a guerrilla war. We can't win. But we can hold on.
— Hold on until when?
— Until people wake up.
— And if they don't wake up?
Solène didn't answer right away.
— Then we'll have lost. But at least we'll have tried.
Étienne came closer. His palm closed on Gabriel's shoulder.
— We need you, he said.
— For what?
— To speak.
— Speak about what?
— About what you've seen. What you know. What you've understood.
— I'm no longer a candidate.
— Exactly. You have nothing left to lose.
Gabriel turned to face the old man.
— You want me to become a ghost too.
— No, said Étienne. I want you to become a voice. A voice that comes from nowhere. A voice they can't control.
— How?
— You speak. You tell people what you've seen. You tell the truth.
— And they'll believe me?
— Maybe. Maybe not. But at least they'll hear.
The screen. The thousands of profiles. The thousands of lives.
He saw his grandfather again. The photograph. The press that told the truth. Then the machine that knew everything. The button he had pressed.
— I don't know if I can, he said.
— Nobody knows, said Étienne. But we have to try.
Gabriel studied the faces around the table. Zara, eyes vacant. Malik, fingers on the keyboard. Solène, pen motionless. Kenza, who hadn't said a word since his arrival.
— I'll think about it, he said.
— Take your time, said Étienne. But don't wait too long.
The screen drew him back in.
— The Parliament of Ghosts won't build itself.
The profiles scrolled past.
Faces. Names. Lives.
People who didn't exist.
But who spoke.
Who argued.
Who changed their minds.
Who formed coalitions.
Who broke them.
People who didn't exist.
But who made the machine doubt itself.
His reflection in the screen overlapped with the images. The campaign. The speeches. The promises. The machine that knew everything. What he might say.
And he remained there, before the screen, while the ghosts argued.