Too Real Not to Be Real

A retiree lost his savings to a face on a video call. The person running the scam was a prisoner. The machine between them was built with American technology. An AP investigation just showed us the factory floor.

Chris Colocousis is not a fool. He is a divorced man in his sixties from Massachusetts who spent a working life saving for a retirement he could count on. Then a woman named Eliza reached out on Facebook. She had a New York number, a job at a well-known financial firm, and a face he could see. When they moved to a video call, there she was, the same blond woman from her photos, right down to the little bags under her eyes.

"She was too real not to be real," he told the Associated Press.

Four hundred thousand dollars later, the retirement is gone. On January 25, 2025, he laid eighty thousand dollars in cash out on his table in neat rows, exactly as the "customer service representative" at his crypto app instructed, because Eliza told him this last payment would finally unlock his money. "You just feel like your whole world fell apart," he said.

I have sat across from people like Chris. The part most people miss is what he said next: those "who think scam victims like him are gullible idiots don't understand the sophistication of criminal organizations behind online fraud." He is right. And a new investigation by the AP and FRONTLINE, reported by Erika Kinetz, just pulled back the curtain on exactly how sophisticated it has become.

I told you this was coming

Back in December I wrote that AI didn't create cybercrime, it just put it on steroids. That was the argument. This is the evidence.

The AP found that American artificial intelligence, chiefly OpenAI's ChatGPT and Google's Gemini, has been built into off-the-shelf scam software with names like KT and 007TG. These tools do not help a con artist write one better email. They let a single worker run fraud like a factory line: automated replies, a role-play chatbot that invents a convincing character, real-time translation across more than a hundred languages, and dashboards that track how many victims each worker is "closing."

One man, Safeer Mohammed Koorimannil, told the AP he chatted with more than a hundred people at once, across dozens of fake profiles, on a single shift. In one month he targeted around fifty thousand people in at least seventeen countries. A widowed tailor. A pastry chef. A sheep farmer. A dairy farmer. Chris, in Massachusetts, was one face in a crowd of tens of thousands, worked by software that never tires and never breaks character.

"Everyone is a robot there," Koorimannil said.

The person scamming you may be a slave

Here is the truth that should sit in your chest for a minute. Koorimannil was not getting rich. He was a prisoner.

He answered an online ad for a tourism job in Thailand. From the Bangkok airport a black car drove him to the Myanmar border, and armed men walked him across a river into a scam compound the United States government has since sanctioned. Supervisors patrolled the desks carrying electric batons. When he was bad at stealing from strangers, he was beaten, his body left "red and swollen with lashes," the AP reported. "When they came near my computer, my hands would shake and sweat," he said. At night he and his best friend shared a narrow bunk, too frightened to sleep alone. It cost his family around five thousand dollars to buy his freedom.

So when we talk about "the scammer," understand that there are often two victims on the line: the retiree losing his life savings, and the trafficked kid being whipped for not stealing fast enough. Both are being fed into the same machine. I write about the human cost of this crime constantly, because it is destroying families on both ends of the wire. This is what that looks like at the source.

Why it feels unstoppable

It feels relentless because it is engineered to be. This is an industry now, and it is enormous. The Federal Trade Commission estimates Americans lost nearly two hundred billion dollars to fraud in 2024. Blockchain analysts at TRM Labs found that the scammers using just one of these toolsets pulled in more than seventy-five million dollars. These are not lone predators. They are organized criminal enterprises, many tied to the same networks I described in my piece on pig butchering, running out of the global cybercrime hubs that treat American retirees as inventory.

And it is about to get worse. "We're moving towards a world where maybe you don't need human scammers anymore," Ari Redbord of TRM Labs told the AP. "All you need is hundreds, thousands, millions of agentic agents who don't need to sleep, don't need to eat, who are 24/7 doing this."

The hardest part is who is carrying the water. The AI models, the internet providers, and the satellite links that make all of this possible are, in large part, American. The companies are not breaking the law. But as Sascha Meinrath of Penn State put it to the AP, "the cost of facilitating scamming is zero." A cybersecurity CEO, Matthew Moynahan, said it should work like a utility: "dirty water shouldn't get to you." Right now, it flows straight to your phone.

What still protects you

I promised you dignity, not fear, so here is the part that matters, and it is the part the criminals hope you skip.

The machine can fake a face. It can clone a voice from a few seconds of audio. It can hold a warm conversation with you in flawless English at three in the morning. What it cannot do is survive a pause.

Every one of these scripts, no matter how advanced the tool behind it, is built to do three things: create urgency, demand secrecy, and move money. Pressure to act now. Insistence that you tell no one. A reason it all comes down to a payment. When you feel those three at once, that feeling is the alarm. It is the point I made in "Deepfakes Are No Longer the Problem. Belief Is." The technology is not the weak link. Your trust is. So protect the trust.

Do not act inside the conversation the scammer controls. Hang up and call back on a number you already have, from your own contacts or the company's real website. Before you move any money, tell one person you trust, out loud. Agree on a family "safe word" now, a phrase only your people know, so that a panicked call in your daughter's cloned voice can be checked in five seconds. And take some heart in this: the same AI being turned against you is being turned back on the criminals. The AP notes that people now use these tools to spot and avoid scams far more often than scammers use them to attack.

Don't carry the shame

The last thing the AP investigation tells victims is the first thing I tell them in an interview room: it is not your fault. The fraudster is to blame. Chris Colocousis was not gullible. He was targeted by an industry that spends millions to build a woman named Eliza who is "too real not to be real," staffed by people who were themselves kidnapped to build her.

You did not fail a test. You were hunted by a factory. The most powerful thing you can do about it is refuse to be quiet. Talk about it at your dinner table. Set the safe word tonight. Tell the story out loud so the next person recognizes the machine before it reaches them. I have written that they didn't hack our computers, they hacked our culture. We take the culture back the same way, one honest conversation at a time.

Source: "Almost Automated: American AI abused for industrial-scale scamming in SE Asia," Associated Press / FRONTLINE, Erika Kinetz, June 29, 2026.