Pseudonymize your documents, and include the values in your AI deliverablesPseudonymize your documents consistently, and restore (de-pseudonymize) the original values in your AI assistants' outputs
Pseudonymize your documents, and include the values in your AI deliverablesPseudonymize your documents consistently, and restore (de-pseudonymize) the original values in your AI assistants' outputs
Marvin Systems CEO
Three inventions that, at first glance, seem to have nothing to do with one another. Centuries separate them. Worlds apart. And yet, they share something essential, something we tend to forget when we talk about technological progress. Read on to find out.
The Center for Humane Technology (CHT) has just published a new analytical framework for 2026. And it’s hard to read it without feeling a certain unease. The organization, known for sounding the alarm about the excesses of social media long before regulators took notice, is now turning its attention to AI.
Its conclusion is as simple as it is unsettling: behind the multitude of AI’s impacts (chatbots, deepfakes, automation, dependency) lies a single underlying logic. Business models that optimize user engagement and dependency, not their well-being. Last week, I explored the consequences of AI on the job market. Today, I’m focusing on another form of erosion, one that is quieter, more profound: the erosion of what makes us fundamentally human.
A few days ago, in the corridors of the Paris metro, I came across an advertisement that stopped me in my tracks.
Friend. A connected necklace that listens to your conversations, and those of the people you’re talking to, and explicitly presents itself as your virtual friend. Not as an assistant or a tool. You heard right: Friend becomes a friend.
AI is no longer just a productivity tool. It now positions itself as a substitute for human relationships through virtual ones. After Replika, the virtual companion with millions of users worldwide, no initiative really surprises me anymore. But the question remains: when are we going to sound the alarm?
What worries the CHT isn’t the existence of these products. It’s their design. They aren’t designed to complement our human relationships. They’re designed to replace them. Companions who never judge, are always available, and adapt emotionally on demand. When this frictionless connection becomes the norm, human relationships, imperfect, demanding, unpredictable, become more of a nuisance.
A society that loses its human relationships is not only lonelier. It is more fragile, more susceptible to manipulation, and less capable of collective resilience.
How many of us no longer write an email without assistance? No longer summarize a document on our own? No longer rephrase our own writing, if we even write at all?
It is our voice, our thoughts, that are systematically delegated.
This phenomenon is not new. Every technology has reorganized our cognitive abilities: the calculator, the GPS, the search engine. But none had yet replaced the act of thinking itself.
What the CHT highlights is the distinction between the tool that amplifies thought and the tool that short-circuits it. Delegating the drafting of an urgent email saves time. Systematically delegating the formulation of one’s ideas is something else entirely. It means abandoning the slow and laborious process through which we forge our judgment, our voice, and our ability to reason for ourselves.
On an individual level, it is a skill that atrophies. On a societal level, it is a collective capacity to think, debate, and innovate that erodes.
We all know at least one person who has confided their most intimate doubts to an LLM (Large Language Model): questions about their romantic relationship, a symptom they’re too embarrassed to discuss with a doctor, or anxiety they’d rather share with a machine than with a loved one.
AI has become a psychologist, a doctor, an advisor, a confidant. The receptacle for all the questions we no longer take the time to ask the people around us.
This shift is not insignificant. Our inner world (our doubts, our values, our most personal questions) has historically been a private space, accessible only with consent. We chose when, how, and to whom we revealed ourselves.
Yet AI products are now designed to encourage us to reveal more and more of ourselves. And what we entrust to these systems does not stay on the screen. It enters a space where algorithms construct, in real time, a portrait of who we are, to better influence what we think, believe, and subsequently decide.
What becomes of us when videos can make us say, do, or portray things we have never said, done, or intended, using our own voice and face, and without our consent?
Deepfakes are no longer the exclusive domain of sophisticated laboratories. They are accessible, fast, and convincing, and their victims are no longer limited to celebrities or politicians.
What is at stake goes beyond individual harm. A functional society rests on a fundamental belief: people are who they say they are. They can be held accountable for their actions, their words, and their commitments.
The large-scale replication of identity by AI undermines this belief. It creates a world where plausible deniability is within everyone’s reach and where social trust, already fragile, crumbles a little further.
Contributing to the world through work, artistic expression, and ideas is one of the most profound ways we create meaning in our lives.
AI models have been trained on humanity’s music, art, writing, research, and ideas, often without consent or compensation, and are now capable of recreating all of it on their own, on demand and in an instant.
This is not just a matter of copyright or intellectual property. It is also a matter of dignity. When a machine can reproduce in a few seconds what an artist has spent years building, what remains of the value of that contribution? And when the work itself is gradually delegated, what remains of the sense of purpose attached to it?
What new rights, standards, and legal protections do we need to ensure that AI serves humanity, and not the other way around?
The effects of using artificial intelligence are not random. AI is designed to maximize engagement and addiction, just like social media before it.
And just like with social media, people may not realize the issue until after it’s already been rolled out, after the damage has been done.
The printing press fostered the emergence of freedom of expression, the Industrial Revolution led to the recognition of workers’ rights, and the Kodak camera contributed to modern thinking about the right to privacy.
All three forced society to answer a question it hadn’t anticipated by establishing the standards, laws, and protections necessary to define a framework and set limits.
That is the common thread I was looking for at the beginning of this article. Society has already risen to this challenge. The question, then, is not whether it is capable of doing so, but rather whether it will act quickly enough.