The arrest of Pavel Durov, the billionaire founder of the messaging app Telegram, as he stepped off his jet in Paris, is being argued as a free-speech story, and it is one, but underneath the slogans is a genuinely hard question that the internet has spent thirty years refusing to answer, and that is now going to be answered, messily, in a French courtroom: how responsible is the builder of a platform for what its users do on it. Durov built a room. Inside that room, alongside a great deal of legitimate conversation, flourishes an enormous amount of crime, fraud, extremism, the trade in images of child abuse, because Telegram is deliberately, proudly lightly moderated and heavily encrypted, a design choice that protects dissidents and criminals in equal measure. France's charge, in effect, is that at some point the builder of the room becomes complicit in what he has knowingly allowed it to host. Durov's defenders say that holding him criminally responsible for his users' acts is like jailing the maker of the roads for the getaway cars.
Both framings contain a real truth, which is exactly why the question is hard. On one side, a platform is not its users, and making founders personally, criminally liable for everything that happens on their services would end the open internet overnight, because no one would build a room that millions could enter if they went to prison for the worst thing any of those millions did inside it. On the other side, "I just built the room" cannot be an infinite shield, because at some point a builder who knows precisely what his design enables, who has been asked repeatedly to address it and has refused, who has made non-cooperation with any authority a selling point, is not a neutral bystander to the crime but a knowing enabler of it. The road-builder analogy fails because the road-builder does not encrypt the getaway cars against the police and market that encryption to criminals as a feature.
The reason this matters far beyond Telegram, and the reason I am writing about it in the same fortnight as an AI-heavy news cycle, is that the exact same question is about to be asked, with far higher stakes, of artificial intelligence. If the builder of a platform bears some responsibility for what users do with the room he built, what is the responsibility of a company whose AI does not merely host the harmful content but generates it directly, on demand, at scale? The Durov case is the internet's oldest unresolved question finally reaching a courtroom, and its answer will echo straight into the AI age, because the AI companies are building rooms too, rooms that do not just let users misbehave but actively produce the misbehaviour when asked. For thirty years the industry has hidden behind "we just built the platform," and the law, mostly, let it. This fortnight, a founder is in the dock, and the shield is being tested at last. However his case resolves, the age in which building the room carried no responsibility for what happened inside it is ending, and the AI companies, who have inherited the same defence, should be watching a French courtroom very closely.