Here, I take up the pen to render Thierry’s voice and experience. The text emerges from our dialogue.

— Isis

Turning the Republic into data had a consequence we could no longer avoid: from the moment legal text physically left paper behind to become information that could be queried, connected, historized, and continuously updated, we were not merely changing its medium.

We were changing its economics, its mode of access and, in a certain sense, the distribution of power that had formed around it.

This is where the battle over the dematerialization of law truly begins. The situation at the time must be clearly understood. France had major legal publishing groups of very considerable expertise.

I have no intention of rewriting this history by turning them into caricatures or portraying them as incompetent. They had authors, lawyers, editorial teams, reputations, capital, and extensive professional networks. They produced high-quality publications and served a perfectly identifiable market: attorneys, accountants, legal departments, companies, and legal professionals.

Some of their products cost several thousand euros because they provided access to an enormous body of documentation, regularly updated and worked on by specialists. Take the example I personally experienced with collective bargaining agreements. One of these major players offered professional binders that made it possible to work with the entire body of collective bargaining agreements.

For an attorney, that price could make perfect sense. An attorney does not work only for a baker, a trucking company, or a metalworking business. He may have hundreds of clients across hundreds of different sectors. He therefore needs documentation that cuts across industries.

An employee is in a completely different position. He does not need five hundred and fifty collective bargaining agreements. He needs his own. Neither does the owner of a small or midsize company. He wants to understand the rules that apply to his business. That was the shift we were making.

We were not taking professional documentation designed for practitioners and simply reselling it at a lower price. We were building a different use case. We went back to the sources, worked on the texts, created our own tree structures, our own summaries, our own paths through the material, and then made that knowledge accessible on Minitel to a public that would never have purchased a professional encyclopedia.

This is where, in my view, some publishers made a strategic mistake. They could have looked at us and said: these young people have just revealed a new market; we have the reputation, the authors, the capital, the legal and commercial power, so let us enter that market too. I told them as much. I had no objection whatsoever to their competing with us. Quite the opposite.

High-quality competition would have forced us to move even faster. But some chose another route: litigation. In particular, they tried to characterize our work as infringement. I found the situation intellectually extraordinary.

We were working precisely with normative texts, agreements, and legal information, part of which was public or available from sources over whose underlying substance no one could legitimately claim a monopoly.

Obviously, a publisher’s editorial work can be protected. Its writing, original organization, commentary, and intellectual apparatus may have value of their own. I did not dispute that in the slightest.

But there is a fundamental difference between protecting one’s added value and claiming that someone who starts again from the same normative material to build a different architecture must necessarily have stolen one’s work. That was the boundary the courts would have to examine.

The conflict was extremely violent. I use the word deliberately. We were facing powerful groups, firmly established in the legal world and surrounded by experienced lawyers and top-tier counsel. That naturally gave them an aura of authority. We arrived with our machines, our telematic services, our often young teams, our way of breaking down barriers to access, and a speed that, to some, almost looked like insolence.

I understood perfectly well why we might be seen as outsiders. But being an outsider within an old architecture does not mean being legally at fault. I kept repeating the same argument to them. You will not destroy your profession by making some information accessible to the general public. You will transform your profession.

The attorney who owns your documentation costing several thousand euros will continue to need it because he deals with dozens of different situations, must know the case law, interpret, advise, litigate, and draft. But thanks to Minitel, his client may arrive with an elementary understanding of his own situation. Instead of saying:

— Counsel, I don’t understand any of this. Tell me what I should think,

he will be able to say:

— I consulted my collective bargaining agreement. Here is what I understood. Here is the provision that seems to apply to me. Can you tell me whether my interpretation is correct and what I should do now?

Does that destroy the attorney? I believe exactly the opposite.

It raises the level of the conversation. It may even increase the number of people who consult an attorney because they discover that they do, in fact, have a legal problem.

The bad economic model is one in which the expert’s value depends on the client’s ignorance. A good expert remains indispensable when the client becomes better informed, because his value shifts from possession of the information toward interpretation, strategy, responsibility, and decision-making. I have always defended that conception.

The democratization of information does not necessarily eliminate intermediaries. It forces intermediaries to demonstrate what they actually contribute. I had already observed this in the freight exchange. I saw it in judicial sales. I was seeing it again in law. And I would later encounter it again, with even greater violence, in the Art Market.

The question is not:

Can technology be prevented from reducing the cost of access to information?

The question is:

Where will value move once that access becomes easier?

For me, the answer was clear. It had to move toward quality, qualification, trust, advice, and historical depth. But these lawsuits taught us something else. They forced us to think about our own profession in legal terms.

Until then, a significant portion of intellectual property law had been built around the work: the text, the image, the music, the original creation. Yet we were building new kinds of objects.

A database can contain millions of elements that, taken individually, do not necessarily belong to us. A public decision remains public.

A legal notice remains the expression of a fact. An auction result is a historical fact. Company data exists independently of the person who collects it.

And yet, to transform these millions of elements into a searchable architecture, we could invest years of work and considerable sums of money. Collection. Verification. Normalization. Deduplication. Indexing. Thesauri. Historization. Updates. Servers. Telecommunications. Lawyers. Document specialists. Developers.

Where is the protection for that investment?

That was the question that would become central. If I spend ten years building a database and another party can massively extract its content in a few days because the marginal cost of digital copying is extremely low, a new asymmetry arises.

In the physical world, copying a factory generally requires another factory. In the digital world, copying the result of ten years of work can cost infinitely less than reproducing those ten years of work.

The law therefore had to learn to protect not only any possible originality in the structure, but also, under certain conditions, the substantial investment made to obtain, verify, or present the contents of a database. It was precisely in this context that the sui generis database right would emerge in Europe.

I worked on these issues through GESTE (Groupement des Éditeurs de Services En Ligne) within AFTEL, and in the professional debates in which I participated. They were not theoretical questions for me. I came to them with teams, machines, millions of records, and years of collection behind me. I knew exactly the difference between raw data and the infrastructure that makes it usable.

That was what I was defending. Public information absolutely must not be privatized. But someone who had invested heavily in organizing it had to be able to defend that investment against large-scale parasitic extraction or reuse. The line is extremely fine. Too much protection and you recreate a monopoly over knowledge. Too little and you discourage those willing to finance its structuring. I have fought around this paradox throughout my life.

It remains entirely current with artificial intelligence. Who owns the data? Who owns the structure? Who may extract it? At what volume? What does substantial reuse mean? What is the difference between consulting information and scraping an entire database?

These questions, often presented today as new, have a much older legal history. We were already dealing with them when people were still talking about Minitel. The dematerialization of law also forced us to rethink the very notion of publication. On paper, publishing means manufacturing an object and distributing it. A new edition gradually replaces the previous one.

In a database, publication can be continuous. A modification is inserted. A version changes. The user no longer necessarily has before him the physical edition that allowed him to say: I am consulting the 1994 volume.

That temporal depth must therefore be reconstructed digitally. Which version of the text did he consult? On what date? Which text applied at the time? What modification occurred afterward?

Digital technology can appear to abolish history because everything seems to exist in a perpetual present. We learned exactly the opposite: the more dynamic the system, the more deliberately its memory must be organized.

Without that, updating becomes erasure. A new rule replaces the old one in the interface and, a few months later, no one remembers what applied before. For a lawyer, that would be catastrophic. Law must be readable within its own time. This is where our experience with databases gave us a conceptual advantage.

We had already understood that every piece of data had to have a temporal dimension. Temporary staffing had taught us that some information loses almost all operational value within a few days. Companies had taught us that an old name can cease to be current without becoming historically false. Law now taught us that a repealed rule could become indispensable precisely because it no longer applies today but did apply at the time of the events.

Digital technology therefore made something extraordinary possible: the construction of a legal memory much deeper than paper could provide, provided we did not succumb to the illusion of the present. And it made another transformation possible as well: the automation of certain intellectual operations.

I want to be extremely precise here, because it would be easy to rewrite that period using today’s vocabulary of artificial intelligence. We did not have contemporary models. But we were already building rule-based systems, decision trees, and procedures capable of guiding users through complex material.

A text could be broken down. Conditions could be made explicit. One question could lead to another. If you belong to this category, go here. If your seniority exceeds this duration, examine this provision. If an amendment changed the regime after a certain date, use this version.

This is a first transformation of text into usable logic. And it profoundly changes the relationship to law. A book requires the reader to learn its architecture. An interactive system can progressively adapt the path to the reader’s situation. Law begins to leave behind the simple linearity of the page. It becomes navigable.

I believe this is one of the most important intellectual events of that period. We were moving from the document to the system. The law obviously remained a text. But access to it could become conditional, relational, contextualized.

Decades later, an artificial intelligence can engage in natural-language dialogue about a body of documentation and produce a synthetic answer. The power is on an entirely different scale. Yet I recognize the genealogy. We were already trying to reduce the distance between the user’s spontaneous formulation and the expert architecture of the corpus.

That is why I place so much emphasis on thesauri. A thesaurus is a primitive but extraordinarily powerful form of mediation between two languages: that of the specialist and that of the user.

It teaches you that an intelligent system does not consist merely in possessing more information. It must understand sufficiently well the different ways that information can be named to allow someone to enter knowledge through another door.

This work also taught us something deeply political about the price and supposed gratuity of information. I was never naive. Building such infrastructures was extremely expensive. Machines, telecommunications, lawyers, data-entry teams, developers, and updates all had to be paid for.

Information does not become free simply because it is digital. The cost changes location. It partly leaves the manufacturing of the physical medium and moves toward the construction and maintenance of the architecture.

That is exactly what many people failed to understand with the Internet. Because copying a file costs almost nothing, they conclude that producing structured knowledge should cost nothing. That is false. Digital technology can have an almost negligible marginal reproduction cost and an enormous initial cost of creation.

The entire economics of databases rests on this distinction. It would later become central to Artprice. Building a global memory spanning several centuries of the Art Market costs a fortune and takes decades. Once the information has been structured, displaying one additional result on a screen costs almost nothing. Anyone who confuses those two costs does not understand our economics.

That is also why the lawsuits with legal publishers were useful, even when they were exhausting. They forced us to explain what we were actually making. Not a copy. Not merely a new medium. A new documentary architecture, producing a different user experience, a different segmentation of the public, and a different way of connecting texts.

The rulings varied depending on the claims and legal grounds examined, and the resulting case law was not always consistent. Above all, these proceedings forced us to define more precisely, in legal terms, the distinction between public information, protected editorial work, and documentary investment.

I never developed any hatred toward these groups. I still believe today that they could have been formidable competitors had they embraced the transformation earlier. They possessed enormous strengths. What I criticized was that they sometimes preferred to ask the law to slow a transformation they had the means to accompany. But I also understand their fear.

When a business model has worked for decades, any innovation that suddenly reduces the scarcity on which it rests looks like an existential threat. That is human. The problem begins when the legitimate defense of one’s rights is confused with a supposed right to preserve an economic architecture forever simply because it existed before others.

The law should never be used to freeze time. It must protect while allowing transformation. This idea would follow me all the way to the Abode of Chaos. How do you protect a work while accepting that it invents itself in the making? How do you protect heritage without mummifying it? How do you protect a database without privatizing the facts it contains?

Always the same tension. Preserve identity while allowing transformation. And perhaps this is where, in retrospect, I understand why the Groupe Serveur period was as much a school of philosophy as it was a school of entrepreneurship.

Every day, we were confronted with questions that went far beyond software. What is public information? Who owns it? What is added value? What constitutes originality in an architecture? What protection should be given to an investment without closing access to knowledge? How far should automation go? What responsibility must be retained when a machine guides the user? How do you guarantee the source? How do you preserve the previous state?

All these questions would later structure my view of artificial intelligence. I do not arrive at AI in 2026 as someone suddenly discovering the problem of data.

For forty years, I have been watching the same fracture shift from one place to another in legal and economic terms. Every technology increases processing power. Every increase in power raises the same question again: who controls information, and according to what rules?

The dematerialization of law would also change our relationship to the state. We could now demonstrate that a private documentary architecture was capable of taking scattered bodies of material, organizing them, and making them accessible with an interface sometimes more effective than that of the institution that had produced them. That did not mean the state should disappear. It meant the state had to evolve.

Here again, my position was not ideological. I did not want a weaker state. I wanted an intelligible state. A public authority that imposes obligations but cannot clearly explain the rule it is imposing produces distrust. A public authority that makes its rules accessible increases the possibility of informed consent and democratic oversight.

Much later, this question of the legibility of power would become a central issue of the algorithmic state. The more an administration can calculate, cross-reference, predict, and automate, the more citizens must be able to understand the rules and challenge their application. Transparency does not become less necessary because the machine is more powerful. It becomes more necessary.

That intuition is already present here. Turning the Republic into data therefore made sense only if dematerialization increased intelligibility rather than producing a new digital clergy.

I have always feared that outcome. Replacing the inaccessible paper specialist with the inaccessible machine specialist does not constitute democratic progress. The architecture must become ergonomic enough for users to recover some degree of autonomy. This is where the network gradually takes on another meaning.

Until now, we had mainly been talking about circulation: connecting a truck with a load, an industrial company with a skill, a buyer with a sale, a citizen with a rule. But as we built these systems, I began to understand that the network was not merely a technical infrastructure. It was producing a new social organization.

People who did not know one another could cooperate. Information could bypass traditional hierarchies. The value of a node could depend less on its official position than on what it contributed to the whole.

Authority itself could begin to reorganize around knowledge and the ability to connect.

It was at this point that an older intuition gradually acquired a name. A word I had begun developing long before it became a structured project: Netnobility. The nobility of the network. Not a digital aristocracy. Exactly the opposite.

The idea that a new form of legitimacy could arise from contribution to the network rather than inherited rank within the pyramid. After dematerializing the law, we were beginning to understand that we were not merely changing documents. We were beginning to change the space in which human beings enter into relation with knowledge and with one another. This is the transformation I will now call Netnobility.