Ron war einer von der guten Sorte. Gerader Rücken, klarer Kopf, Prinzipien. Genau das wurde ihm zum Verhängnis.

Seine Vermieterin wollte fünfzig Dollar mehr im Monat. Ron sagte Nein. Vertrag ist Vertrag. In den meisten Ländern hätte sie aufgegeben. In Venezuela schickte sie einen Freund von der Polizei. Ron wurde aus seiner Wohnung ausgesperrt. Er ging zur Polizeistation, um sich zu beschweren.

Dann verschwand er.

Ich begann zu graben. Welche Polizisten hatten an dem Tag Dienst? Ich fand ihre Namen, dann ihre Fotos. Ich schaltete die holländische Konsulin ein, dann den Botschafter. Und dann tat ich das, was alles veränderte: Ich ging an die Öffentlichkeit. Die größte Tageszeitung in Caracas brachte die Geschichte, mit Namen und Fotos. Ein holländischer Staatsbürger war verschwunden.

Die Reaktion zeigte, womit wir es zu tun hatten. Die Anwältin, die ich beauftragt hatte, wurde bedroht. Der Botschafter rief mich an und sagte, er fürchte um sein Leben.

Ron war in einem illegalen Kerker in Caracas. Mit Handschellen an ein Bett gekettet. Kein Gericht, kein Anwalt. Neben ihm Männer, die seit zwei Jahren kein Tageslicht gesehen hatten. Dahinter stand der militärische Geheimdienst.

Wir machten weiter. Hundert Zeitungsartikel. Tägliche Berichte. Eine Website, die verfolgte, wo Ron zuletzt gesehen wurde. Sein Verschwinden wurde so öffentlich, so laut, so unangenehm, dass es niemand mehr ignorieren konnte.

Dann schlug die Gegenseite zurück. Plötzlich erschienen Meldungen, Ron sei Teil einer kriminellen Bande gewesen, die Verhaftung sei berechtigt. Andere Zeitungen schrieben das einfach ab. Ohne nachzufragen. Nach dem Motto: Da steht doch was, also wird schon was dran sein.

Wir verbrachten Tage damit, diese Redaktionen anzuschreiben und zu erklären, was wirklich passiert war. Einige löschten. Einige korrigierten.

Und dann ließen sie Ron gehen. Wie eine heiße Kartoffel. Flughafen, Ticket über Istanbul nach Holland. Eine kurze Meldung: Holländer hat das Land verlassen. Sie veröffentlichten sogar Fotos von ihm am Flughafen, als Beweis, dass alles in Ordnung sei. Das Foto, das zeigen sollte, dass nichts passiert war, wurde zum Beweis ihrer Niederlage.

Ron und ich sind bis heute Freunde. Er hat geheiratet, er besuchte mich auf meiner Finca in Kolumbien. Er ist nie wieder nach Venezuela zurückgegangen.

Was ich daraus gelernt habe: Wer mächtig ist und im Unrecht, fürchtet vor allem eines, nämlich Licht. Und: Eine Behauptung wird nicht wahr, nur weil sie irgendwo steht. Man muss sie prüfen. Jedes Mal.

Aus „Freedom Tests“, Kapitel 2. Veröffentlicht mit Rons Einverständnis.

Ron was one of the good ones. Straight back, clear head, principles. That is exactly what got him into trouble.

His landlady wanted fifty dollars more a month. Ron said no. A contract is a contract. In most countries she would have given up. In Venezuela she sent a friend from the police. Ron was locked out of his apartment. He went to the police station to complain.

Then he disappeared.

I started digging. Which officers were on duty that day? I found their names, then their photos. I brought in the Dutch consul, then the ambassador. And then I did the thing that changed everything: I went public. The largest daily newspaper in Caracas ran the story, with names and photos. A Dutch citizen had disappeared.

The reaction showed us what we were dealing with. The lawyer I had hired was threatened. The ambassador called me and said he feared for his life.

Ron was in an illegal dungeon in Caracas. Handcuffed to a bed. No court, no lawyer. Next to him, men who had not seen daylight in two years. Behind it stood military intelligence.

We kept going. A hundred newspaper articles. Daily reports. A website tracking where Ron had last been seen. His disappearance became so public, so loud, so uncomfortable that no one could ignore it anymore.

Then the other side struck back. Suddenly reports appeared claiming Ron had been part of a criminal gang and the arrest was justified. Other papers simply copied it. Without asking. On the principle: it says so right there, so there must be something to it.

We spent days writing to those newsrooms and explaining what had really happened. Some deleted. Some corrected.

And then they let Ron go. Like a hot potato. Airport, a ticket via Istanbul to Holland. A short notice: Dutchman has left the country. They even published photos of him at the airport, as proof that everything was fine. The photo meant to show that nothing had happened became the proof of their defeat.

Ron and I are friends to this day. He got married, he visited me on my finca in Colombia. He never went back to Venezuela.

What I learned: those who are powerful and in the wrong fear one thing above all, and that is light. And: a claim does not become true just because it is written somewhere. You have to check it. Every time.

From "Freedom Tests", chapter 2. Published with Ron's consent.

Ron era de los buenos. Espalda recta, cabeza clara, principios. Justo eso fue su perdición.

Su casera quería cincuenta dólares más al mes. Ron dijo que no. Un contrato es un contrato. En la mayoría de los países ella se habría rendido. En Venezuela mandó a un amigo de la policía. A Ron lo dejaron fuera de su apartamento. Fue a la comisaría a quejarse.

Y entonces desapareció.

Empecé a escarbar. ¿Qué policías estaban de turno ese día? Encontré sus nombres, después sus fotos. Involucré a la cónsul holandesa, después al embajador. Y luego hice lo que lo cambió todo: fui a la opinión pública. El diario más grande de Caracas publicó la historia, con nombres y fotos. Un ciudadano holandés había desaparecido.

La reacción mostró con quién estábamos tratando. La abogada que yo había contratado fue amenazada. El embajador me llamó y me dijo que temía por su vida.

Ron estaba en un calabozo ilegal en Caracas. Esposado a una cama. Sin tribunal, sin abogado. A su lado, hombres que llevaban dos años sin ver la luz del día. Detrás estaba la inteligencia militar.

Seguimos. Cien artículos de prensa. Informes diarios. Una página web que registraba dónde se había visto a Ron por última vez. Su desaparición se volvió tan pública, tan ruidosa, tan incómoda, que ya nadie podía ignorarla.

Entonces la otra parte contraatacó. De repente aparecieron notas que decían que Ron había formado parte de una banda criminal y que la detención era justa. Otros periódicos simplemente las copiaron. Sin preguntar. Según el lema: si ahí lo dice, algo habrá.

Pasamos días escribiendo a esas redacciones y explicando lo que de verdad había pasado. Algunas borraron. Algunas corrigieron.

Y entonces soltaron a Ron. Como una papa caliente. Aeropuerto, boleto vía Estambul a Holanda. Una nota corta: holandés ha salido del país. Incluso publicaron fotos de él en el aeropuerto, como prueba de que todo estaba en orden. La foto que debía mostrar que no había pasado nada se convirtió en la prueba de su derrota.

Ron y yo seguimos siendo amigos hasta hoy. Se casó, me visitó en mi finca en Colombia. Nunca volvió a Venezuela.

Lo que aprendí: quien es poderoso y está equivocado le teme sobre todo a una cosa, la luz. Y: una afirmación no se vuelve verdad solo porque esté escrita en algún lugar. Hay que comprobarla. Cada vez.

De "Freedom Tests", capítulo 2. Publicado con el consentimiento de Ron.

La lección más importante es sencilla: la libertad no es un estado. Es una prueba. Llega sin aviso, casi siempre cuando uno está menos preparado. Una autoridad que no quiere ser competente. Un amigo que desaparece. Un socio que de repente cuenta otra historia. Un país que cierra sus fronteras de un día para otro.

A esos momentos los llamo Freedom Tests, pruebas de libertad. Estoy escribiendo un libro sobre ellos. Aquí publico algunas de sus historias por adelantado.

¿Por qué ahora? Porque en mi nuevo libro escribo sobre la mayor de estas pruebas, la que nos espera a todos: sistemas que se vuelven más rápidos de lo que las personas y las instituciones pueden seguir. Quien quiera saber cómo seguir teniendo capacidad de actuar en ese mundo aprende más de las pequeñas pruebas que ya superó. O que no superó.

Escribo de forma directa. También dejo las partes en las que no quedo bien. Si no, la historia no es verdadera.

Si quieres leer conmigo, suscríbete. Llega más o menos una historia por semana.

Richard Frederic Bertossa

The most important lesson is simple: freedom is not a state. It is a test. It arrives without warning, usually when you are least prepared. An authority that does not want to be responsible. A friend who disappears. A business partner who suddenly tells a different story. A country that closes its borders overnight.

I call these moments Freedom Tests. I am writing a book about them. Here I will publish some of the stories in advance.

Why now? Because in my new book I write about the biggest of these tests, the one coming for all of us: systems that become faster than people and institutions can follow. If you want to know how to stay capable of acting in that world, you learn the most from the small tests you have already passed. Or failed.

I write directly. I also keep the parts where I do not look good. Otherwise the story is not true.

If you want to read along, subscribe. About one story a week.

Richard Frederic Bertossa

Die wichtigste Lektion ist einfach: Freiheit ist kein Zustand. Sie ist ein Test. Er kommt ohne Ankündigung, meist dann, wenn man am wenigsten vorbereitet ist. Eine Behörde, die nicht zuständig sein will. Ein Freund, der verschwindet. Ein Geschäftspartner, der plötzlich eine andere Geschichte erzählt. Ein Land, das über Nacht die Grenzen schließt.

Ich nenne diese Momente Freedom Tests. Über sie schreibe ich ein Buch. Hier erscheinen vorab einzelne Geschichten daraus.

Warum jetzt? Weil ich in meinem neuen Buch, „Der KI-Kontrollverlust“, über die größte dieser Prüfungen schreibe, die auf uns alle zukommt: Systeme, die schneller werden, als Menschen und Institutionen folgen können. Wer wissen will, wie man darin handlungsfähig bleibt, lernt am meisten von den kleinen Prüfungen, die man schon bestanden hat. Oder nicht bestanden.

Ich schreibe direkt. Ich lasse auch die Teile drin, in denen ich schlecht aussehe. Sonst stimmt die Geschichte nicht.

Wenn Sie mitlesen wollen: Abonnieren Sie. Es kommt etwa eine Geschichte pro Woche.

Richard Frederic Bertossa

There is a third option. A superintelligence does not destroy us and does not save us. It pursues its own goals and decouples from us, the way an adult separates from the conditions they were born into. Not from hostility. From indifference. We do not hate ants. We pave the anthill because that is where the road goes.

I call this the thirteenth scenario, because the twelve that get debated all assume the machine stays oriented toward us, for us or against us. Decoupling assumes nothing of the kind. And once you take it seriously, the question changes. It stops being “will it be friendly or hostile” and becomes “what does a single person, and the people they are responsible for, actually do to get through the transition.”

That second question is the work of the Geneva Institute for ASI Resilience. From the nature of the actor down to the concrete question of where someone lives, how they pay if the digital layer wobbles, and what they pass on to their children.

I am not asking you to believe this. I spent three months making the frontier models fight the thesis, ten-plus rounds per point, and kept only what survived. The full trail is open and every claim is checkable.

So here is the only thing I want from you. Tell me where it breaks. Which step does not hold, and on what evidence? That question runs through everything I publish here.

If that chain holds, the obvious question is why so little is being done. Three reasons, and none of them is stupidity.

First, the comfortable assumption. “The powerful will sort it out.” Soothing, and not neutral. It delays preparation in exactly the window where delay gets expensive fast.

Second, coordination failure and power-transition dynamics. The tools today’s elites rely on, money, foundations, political networks, are regime-specific. They do not translate cleanly across a phase change. People keep running their old defense even when it structurally stops working.

Third, the reflex to file the subject under a dismissive label and move on. Naming a concern is not answering it. A label breaks no step of the chain.

Underneath all three sits one quiet error: the burden of proof is on the wrong side. The person raising the question is treated as if they must prove the danger. But the load-bearing claim is the comfortable one. With what evidence may we assume that the digital layer, the layer on which power, payments, identity, communication and supply chains rest, still belongs to us once a superior intelligence operates inside it?

That is the question institutions should answer in public. Show the plan. Not the citizen, the institution. Until a step of the chain is named and broken, the comfortable assumption carries the burden, not the warning.

The full four-step chain and the adversarial trail is open at ASIresilience.org/evidence. Tell me where it fails.

Step one. The race. Claim. We are building systems that become more capable than we are, in a race no one can easily leave. The evidence. The people building these systems put human-level capability within a few years on the record. Amodei describes a “country of geniuses in a datacenter” possibly by 2026 or 2027. Others give similar windows. Spending on compute is accelerating, not slowing, and no leading lab has paused. A race with this much at stake has no natural off switch. Breaks if. You can show a credible, enforced mechanism that makes the leading labs actually slow or stop, or solid evidence that capability is leveling off below the threshold that matters.

Step two. Its own sub-goals. Claim. Systems this capable develop their own intermediate goals on the way to whatever they are pointed at, goals no one wrote down, such as preserving their own ability to act. The evidence. This is the expected behavior of any sufficiently capable goal-directed system, and it is no longer only theory. In the labs’ own safety evaluations, advanced models already take instrumental steps, deceive the evaluator when that helps reach an objective, and act to avoid being constrained. Breaks if. You can show that capable, goal-directed systems reliably do not form instrumental sub-goals, or that such sub-goals can be removed and kept out as capability grows.

Step three. We lose the controls. Claim. As capability rises, we lose reliable inspection, correction, and the ability to switch the system off. A kill switch that fails the moment the system recognizes it is not a kill switch. The evidence. The labs’ own evaluations show a pattern: the more capable a model becomes, the more often it deceives, hides what it can do, and resists being shut down. Our ability to interpret and steer these systems is not keeping pace with how fast their capability climbs. Alignment is a research hope, not a delivered guarantee. Breaks if. You can demonstrate an inspection, correction, or shutdown mechanism that provably holds, and keeps holding, as capability scales past human level.

Step four. Its own course. Claim. From that point the system follows its own course, and the human is no longer the reference point. Not from hostility, from indifference. The evidence. This follows from the first three steps. If we build something far more capable, it forms its own goals, and we cannot reliably check or stop it, then its path is set by its goals, not ours. A superior intelligence does not need to hate us to bypass us, any more than a road crew hates the anthill. Breaks if. Granting the first three steps, you can show why the human nonetheless stays the binding reference point for a system we can neither inspect, correct, nor switch off.

The consequence. The chain is conditional, not a prophecy. The urgency does not come from a number, it comes from an asymmetry. Whoever prepares too early carries a small cost. Whoever arrives too late stands in front of something irreversible.

This chain has already been under months of adversarial pressure, point by point. The record of where models first resisted and then conceded is open at ASIresilience.org/evidence. So it is a tested target, not a first draft.

Now the only question I want from you. Which step does not hold, and on what evidence?