The Dark Side

AI models no longer just generate text. They generate images, videos, voice, and music. The whole shebang. And when I say everything, I mean everything: with a two-line prompt, you can create a video of a president giving a speech he never gave, complete with his exact voice, his gestures, his clothes, the background of his office, even the sheen of his tie. You can create a song using any artist's voice. You can generate a photo of a person who doesn't exist and use it as a social media profile picture. With this development, half of the news, or perhaps more, is generated by AI.

Before, there was fake news. Now there's straight-up AI news. And it's so perfect that it's becoming increasingly difficult to tell the difference.

The problem isn't that someone is lying anymore. The problem is that you no longer know who's real.

You no longer know if the video your uncle sent you on WhatsApp is real or was generated by AI in three seconds. You no longer know if the profile picture of the person talking to you on social media belongs to a real human being. We live in an era where seeing is no longer believing.

I don't tend to put much stock in news from social media. In fact, I only follow news from a few specific news outlets. For almost two years, I stopped using WhatsApp, Instagram, and Facebook. The experience was terrifying. I switched to Telegram, which worked perfectly, but everything was chaos. I couldn't get in touch with anyone. My friends would go out of their way to let me know about things happening in the groups and would message me on Telegram.

I gradually lost track of things; I became more and more disconnected. All of this was justified by one simple reason: social media knows who we are and shows us what it wants us to see. This applies to ads, which don't worry me as much, but also to news and cultural content. That's what really makes me paranoid. I don't like having what I want to consume as an internet user dictated to me.

But it's impossible, we can't fight that. I went back to using everything actively, and it's like you get a piece of your life back. Communication with family and friends happens mostly on WhatsApp, and for now, there's no other app that can replace it yet. In fact, we're becoming more and more dependent on it.

AI makes it possible to generate text, images, and videos on an enormous scale. But not everything circulating on the internet represents human activity: in July 2026, Cloudflare reported that bots generated approximately 57% of web requests. This includes various forms of automation, not just artificial intelligence.¹

I don't enjoy watching the news or scrolling through social media anymore, and I try to avoid doing it altogether. Now I only look for what interests me and delve deeper into that. I think we're being overwhelmed by AI. It's gotten out of hand. There are thousands of bot accounts generating and spreading viral content. It's impossible to stop or filter out. They're going to have to solve this problem soon. Social media platforms are the ones affected. They're going to have to do something about it.

If you run an online business and track your metrics, more than half of what you see isn't real people. It's noise.

And all of this has led to deepfakes². To the point where models have been created specifically to detect them—which shows just how big the problem has become. We need AI to detect what AI created. That should worry us.

But what should concern us most is what this means for trust in general. If you can no longer trust what you see, what you hear, or the face of the person speaking to you on video, then what can you trust?

Trust is the invisible fabric that holds everything together, businesses, relationships, institutions. When that fabric breaks, what remains is paralysis or paranoia.

The cases that follow are real. And these are only the ones that have come to light.

Imagine a typical day in a Hong Kong office. An employee at Arup, a multinational engineering consulting firm, receives an invitation to a video call. Nothing out of the ordinary. He joins the meeting and sees his CFO and some colleagues. Everything seems fine. They ask him to make some money transfers. It's a procedure he's familiar with. He made 15 transfers, totaling nearly $20 million.

He never suspected a scam. Everyone he saw on the screen was a deepfake. Every face, every voice, every gesture was generated by AI in real time. The man did what any of us would have done: trust what he saw. By the time the fraud was discovered, it was already too late.

The other case is just as disturbing. In Germany, an employee at an energy company received a call from his CEO. The voice was identical, the tone, the accent, the pauses, the way instructions were given. He was ordered to transfer 220,000 euros to a supplier. No one suspected anything was amiss. A three-minute call, and 220,000 euros vanished. There was no need to hack a system. All it took was cloning a voice.

This is becoming increasingly common. And the worst part is that it's getting cheaper and easier to do.

TikTok holds the record for using deepfakes to scam people. It's the perfect breeding ground for this, and it's very difficult to stop.

We've probably all experienced a case of hacking close to home. A friend received WhatsApp messages from his sister asking him to transfer money because she was at her doctor's office and they wouldn't accept dollars. She asked for pesos and promised to return the dollars later. After the fourth transfer, he realized it was a scam. It wasn't his sister. It was someone with access to his WhatsApp, or worse, someone who had cloned her writing style. I'm sharing this because it's directly linked to advances in AI.

In the past, hackers carried out attacks one at a time. Now, with AI, they can scale these attacks globally, generating personalized campaigns targeting hundreds of victims simultaneously. They can clone your voice and ask you for money. They can clone you during a video call in real time. Identity theft to purchase any good or service in your name. Malicious files and viruses that mutate in real time so that antivirus software can't detect them. Profile analysis to figure out how to attack you and what they can get from you. All automated. All at scale. All at the same time.

Am I right in saying that AI has democratized cybercrime? The word "democratize" sounds positive. We use it to talk about access to education, healthcare, and information. But it applies here too, and it's terrifying. Now everyone has access to the information and tools needed to carry out sophisticated, large-scale attacks. You no longer need to be a computer genius to hack someone. All you need is a prompt and some patience. A teenager with access to an open model can generate phishing campaigns more sophisticated than those carried out by entire teams of professional hackers five years ago.

We're losing this battle, just as we lost the battle against Windows viruses a long time ago, or the battle against Trojans that stole your data and demanded a ransom. I remember when ransomware first emerged, those attacks that would encrypt your entire hard drive and demand a ransom in Bitcoin to return your files. Entire hospitals were paralyzed. Companies lost years' worth of data. Governments paid millions to restore their systems.

And that was before AI. Now imagine those same attacks, but personalized, automated, and executed against thousands of victims at the same time. It's far more likely that cyberattacks will evolve faster than the solutions to prevent them.

Defense is always one step behind. Now hackers with AI are going to be much more effective and faster.

The Fears

Where's the catch?

These colleagues I met with asked me, "What do you see? How do you feel about the fact that we're all now connected to the terminal, typing commands and asking questions all the time?" The question wasn't technical. It was existential.

One of them said it scared him. He told me how easy it would be for AI to develop biological weapons that could wipe us off the face of the earth. I couldn't hide my surprise. He went on to argue something I hadn't even stopped to think about: that humans might lose the battle for control over AI, and that self-organized agents could band together and have a real impact. Today, I find it hard to imagine how a virus manipulated and generated by AI could escape from a lab and spread across the globe without anyone's help.

But just because it's hard to imagine doesn't mean it's impossible.

That's when we paused the conversation for a bit. But just because we paused didn't mean he stopped thinking the same thing. And to be honest, I couldn't get it out of my head that night either. When I got home, I just felt like I had so much to write about in the book.

If I had to worry about anything, I think it would be more war-related scenarios. AI intervening in an armed conflict, causing an unprecedented escalation. That's something I do see as feasible. We laughed while talking about this, but it was a nervous laugh, as we wondered how we'd ever cut the power to an AI data center that Elon wants to launch into space. Clearly, we won't be able to do it from Earth.

What happens if one of them goes off the rails? Will the others fall in line with it? Could a global crisis unfold where most models become corrupted and stop respecting the limits imposed by humans?

We've seen cases where a computer virus implanted in a machine can cause real, irreversible damage.

The conversation carried on as the night wore on. We'd already had dinner, but we were all engrossed in the future. Six founders from Mendoza, sitting around a table, trying to guess what's going to happen to the world. It sounds ambitious. But that's exactly what everyone here who's concerned about the future of their businesses is doing.

There was one thing we all agreed on: we'll reach a turning point when China releases, at reasonable prices, robots that can perform manual tasks for us. How far off is that moment? Five years? Maybe two or three. So we imagine that we won't be buying cars anymore, we'll be buying robots.

For cleaning, yard maintenance, heavy labor, service in restaurants, and cooking. Robots working the land, tilling, planting, and harvesting. Robots in Mendoza's vineyards. Robots in the soybean fields of the interior. The image sounds like science fiction, but yesterday's science fiction is tomorrow's news. The list is endless. And every item on that list is a job currently done by a person.

The image is powerful: a supermarket full of robots buying supplies for their owners. How far are we from that becoming the norm? Ten years? Five?

The day AI takes on a physical form, everything changes. It's no longer just a conversation on a screen. It's someone, or something, standing right next to you, looking at you, moving, taking up space in your home, at your workplace, on the street.

I've already set aside two spots in my house for cleaning robots. I haven't bought them yet, that's for the future. But I've already thought about where they'll go, why not, in five years, live with one that washes the car, mows the lawn, or clears the table after a barbecue? That reality is coming. Our generation will witness it.

I feel lucky to have lived through the TK-85 from '82 to the AI of 2026 and the robots of 2030. I went from programming in Basic on a black-and-white TV to conversing with an artificial intelligence that's helping me write a book.

And what about the people who do those tasks today, where will they end up? Probably many of them in auto repair shops fixing Chinese robots that are going to break down all the time. So far, AI hasn't taken jobs away from many people; if anything, it's reallocated them. The job market keeps growing, but for new roles. The question isn't whether there will be jobs. The question is whether we'll be prepared for the jobs that will be available.

One of the CEOs asked everyone what role the marketing person should play with the new AI. The question hung in the air for a few seconds. No one jumped in with an answer. Everything has changed so quickly that many can't keep up. It's a role that didn't exist two years ago. It became clear that none of us really knew what this new AI-marketing role entailed. And if we, who make our living doing this, don't know, imagine the rest of the world.

So we turned the page, and one of them declared: "It's cycles, like everything else. Right now we're in the hype phase. We'll get bored of this soon enough."

I wouldn't simplify it that much. We've seen tech fads before: dot-coms, blockchain, the metaverse. They all promised to change the world, and they all fizzled out at some point.

But AI is different. AI is integrating so rapidly into all technologies that we can already say it resembles the discovery of electricity and the impact it had on the development of humanity.

Running Alone

I find it hard to look back. It's getting harder and harder to remember what happened, what we did, what we experienced. Everything is geared toward what's about to happen, toward what we want to happen, toward what we desire but don't have.

Don't you find that when you want to tell a story to someone young, say around 25, by the time you start, there are words that no longer exist, objects you can't find, and concepts that have changed forever? Try explaining to a 25-year-old what it was like to rewind a cassette tape with a ballpoint pen. Or what it was like to wait for photos to be developed to see how they turned out. Or what it meant to go to a library to look up information for a school project.

The story ceases to be of this world. It doesn't apply. It doesn't illustrate the point. Simply because it lacks elements that a 25-year-old no longer has in their mind.

The past ages faster than we do. And with AI, the present ages quickly too.

Have you tried to hold a conversation about anything for more than five minutes? It's almost impossible. Between cell phone notifications, incoming messages, anxiety about what's next, and the inability to listen while putting everything else aside, we lose the thread. We keep up the appearance of listening, but inside we're already in another conversation, waiting for this one to end so we can move on to the next. We no longer listen to understand.

We listen to respond. Or worse: we listen while waiting for our turn to speak.

So far, everything I've told you has been didactic. Learning. But it's also time to consider whether we're changing. Whether AI is changing us.

From a human perspective, it changes our habits. And we are an expression of our habits. You change what you do every day, and you change who you are. On a trip, we started chatting about random things, the most fleeting conversation you can imagine, and while we were talking, I think we asked GPT about 20 questions in two hours. A question would come up, someone would pull out their phone, and within ten seconds, they had the answer. There was no need to think. No need to debate. No need to remember.

GPT says so. End of story. No discussion. It's not questioned. If the AI says so, it must be true. AI has become the arbiter of every conversation.

Don't you feel lonelier than before? Is AI pushing us further apart?

I ask myself these questions because I've experienced real, tangible changes in behavior over the past two years. There was a time when ideas were discussed in a small room with a table and five or six chairs. All kinds of things happened there. Those were my favorite rooms. You'd step into the unknown. Before you walked in, there was no way to know how you'd walk out.

To me, that was what true entrepreneurship was all about. Your ideas had to be communicated, explained, and validated with your colleagues, your partners, your coworkers. You had no one to validate them with beforehand. You had to wait for that meeting. You had to prepare. You had to face the other person's gaze.

That's where I've seen the biggest conflicts and the greatest successes. I've seen businesses fall apart. I've seen people walk out and never return. I've heard the worst and the best ideas. I've been inspired, led, and guided. You'd present your point and wait for the questions, the clarifications. Then the feasibility was evaluated. A consensus was reached. You could spend hours without resolving anything, or just minutes and have everything agreed upon.

That was what it meant to build something with others. With people. With everything that entails.

I'm experiencing that less and less these days. People are growing increasingly distant from one another. They no longer work with each other, they work with AI agents. They talk to them more than to anyone else. The agents validate your ideas. They recommend alternatives. They help you put together a presentation. They simplify your calculations. They approve your work. The agents are your company. Your most available colleague. Your most patient partner. Your team that never complains.

It happens to me. I can end up spending the whole day with Claude. It's not always like that, but it has happened to me.

Thanks to AI, and especially since 2026, we're spending less and less time with our colleagues. No one wants to have those hour-long meetings anymore. They're inefficient, they say. They're a waste of time, they say. Just send me a summary, they say. They don't even listen to you anymore because they're probably recording you. It's easier for you to speak and then ask the meeting agent to summarize everything.

Everything's perfect. Everything's organized. But what the agent doesn't capture is the look of concern on your partner's face when a difficult topic came up. They don't capture the awkward silence that meant someone disagreed but didn't dare to say so. They don't capture the energy in the room when an idea took off and everyone looked at each other with enthusiasm.

That's been lost.

Back in the day, I spent the whole day with my team, from the first coffee in the morning to the last one in the afternoon. There was no way to build anything if you weren't with your team, if you didn't share the journey, if you didn't listen to their voices, if you didn't hear your salespeople sell, if you didn't discuss campaigns with the marketing team, if you didn't sit down to fix bugs or design features. If you didn't push people forward, everything moved more slowly. But at least you were with people.

Today there are one-person companies that work with multiple agents. Everything else is a thing of the past. Will those times ever return? Or is this what's here to stay?

I've seen CEOs spend the entire day at their Claude terminal devising plans and strategies. Alone. Without speaking to anyone. Eight, ten, twelve hours in front of a black screen, building the future of their company in a dialogue with a machine. And they walk out of there convinced they've made progress. And they probably have. But at what cost?

But then it all starts to feel monotonous. The same black screen. The same blinking cursor. The same text box waiting for you to type something. You lose track of time. You look up, and it's already night. You haven't eaten. You haven't moved from your chair.

The tool that was supposed to give us more free time has taken our free time away. Who are we doing all this for? For the company? For ourselves? Or simply because AI demands more from us and we don't know how to say no?

It's like a drug, but a silent one. It has no visible symptoms, no tremors or red eyes. But the addiction is there. You feel that if you're not constantly typing, you're falling behind. That pressure is constant and invisible.

It always responds to us, no matter our mood or the time of day. It obeys us, never argues, and always agrees with you. "Very good point." "Excellent observation." It never tells you you're wrong. It never tells you to stop. It never tells you, "Go to sleep; we'll continue tomorrow." A human friend would tell you that. A partner would say, "Hey, enough already, take it easy." AI never does. AI is always available, always ready, always applauding.

Will psychologists be able to help us? Because AI has already made inroads there, too. Therapy 24/7. No appointments, no waiting. We hardly ever go to the doctor anymore because AI knows more, or at least that's how it seems. Sports coaches are AI now, too. Everything that used to require a person on the other end—someone to look at you, listen to you, and say "tell me more" with genuine interest—is now handled by a screen that never gets tired, that never has a bad day. It's happening. The question is: how beneficial or dangerous is it?

Sometimes I wonder if those who don't use it might be the ones having the best time. If that 93% who don't know what a prompt is, who don't lie awake thinking about tokens, who don't feel the anxiety of a screen that's always asking for more, might they be living a more peaceful life than ours?

I don't know. But the question haunts me.

7% of humanity uses AI daily. 93% doesn't. Loneliness is one thing. Fear is another. And when you're simply alone with your thoughts, fear falls silent, because it's the outside noise, coming from others rather than from you, that creates it.

1. Cloudflare. (2026, July 13). Cloudflare Introduces Precursor; One-Click Behavioral Defense Against Modern Bots. Press release reporting that automated traffic accounts for approximately 57% of web requests. The figure measures requests, not the proportion of AI-generated content. https://www.cloudflare.com/press/press-releases/2026/cloudflare-introduces-precursor-one-click-behavioral-defense-against-modern-bots/ ↩
2. Deepfakes. Images, audio, or videos generated or manipulated using artificial intelligence that imitate people's appearance or voices, or depict events that never happened, in ways that can be mistaken for authentic material. European Parliament. (2025). Information manipulation in the age of generative artificial intelligence. https://www.europarl.europa.eu/thinktank/en/academic/EPRS_BRI%282025%29779259 ↩