Everyone is a robot there, he said, speaking of the scam compound.
From lawless compounds in Myanmar, a vast and industrialized fraud network reaches into the lives of ordinary people across 17 countries — enabled not by shadowy foreign technology, but by the familiar tools of American innovation: AI models, satellite internet, and cloud infrastructure. An investigation by the Associated Press and PBS Frontline reveals how the digital supply chain, from ChatGPT to Starlink, has become the invisible backbone of a criminal enterprise that traffics human beings, strips victims of their life savings, and operates with near-total impunity. In the absence of legal accountability, the machinery continues to expand — and the human cost, measured in blindness, death, and shattered retirements, grows with it.
- AI software built on ChatGPT and Gemini allows a single trafficked worker to simultaneously target 50,000 people across dozens of languages, turning fraud into an automated industrial process.
- One in five internet connections traced to Myanmar scam compounds routes through American companies — including AT&T, Starlink, and Oracle — making U.S. infrastructure a structural partner in the crime.
- Trafficked workers face electric batons, torture, and death when they fail to meet quotas; one man lost his sight permanently, another disappeared after collapsing and was never seen again.
- A Massachusetts man lost $400,000 in retirement savings to a romance scam so convincing it included a video call, a New York phone number, and a cash pickup at his door.
- Despite a high-profile crackdown in late 2025, scammers simply relocated — satellite imagery confirms at least 25 new compounds built since, with Starlink connections active at 13 of them by spring 2026.
- Tech companies cite the absence of legal penalties as the reason they do little more than the minimum, while the UK, EU, and Australia move toward mandatory accountability frameworks that the U.S. has yet to adopt.
American technology has become the engine of an industrial-scale fraud operation stretching across continents. A joint investigation by the Associated Press and PBS Frontline found that AI models from OpenAI and Google, satellite internet from Starlink, and cloud infrastructure from U.S. providers are powering organized crime networks operating from Myanmar — networks that reach victims in at least 17 countries and have trafficked an estimated 300,000 people.
The machinery is both sophisticated and brutal. Safeer Mohammed Koorimannil, trafficked to a Myanmar scam compound, spent his days posing as a young Singaporean woman, managing over 100 simultaneous conversations using software called KT — built on American AI — that generated replies, handled translation across dozens of languages, and tracked his performance metrics. In a single month he targeted roughly 50,000 people. When his numbers fell short, supervisors beat him with electric batons. A similar platform, 007TG, received $860,000 in cryptocurrency payments over 20 months; the scammers using it earned at least $75 million.
The human wreckage is not confined to those trafficked. An Ethiopian engineer named Ebisa was beaten so severely at a compound near the Thai border that he lost sight in one eye — and by the time he reached home, doctors told him it was too late to restore it. A Nigerian man named Obinna Okeadu collapsed after a punishment session and was taken away; his name was deleted from work chats the next day, and he was never seen again.
Victims on the other end of the screen face their own devastation. A divorced man in his 60s from Massachusetts was drawn into a romance scam so convincing it included a video call and a New York phone number. He lost $400,000 — his retirement — and now sometimes struggles to leave his house.
Analysis of more than 200,000 device connections at four Myanmar scam compounds found that one in five signals was carried by a U.S.-registered company. Starlink emerged as the dominant internet provider, even to known scam centers, despite congressional pressure and a publicized crackdown in late 2025. When authorities demolished a high-profile compound and seized Starlink terminals, the scammers scattered — and brought their equipment with them. Satellite imagery confirmed at least 25 new compounds built since the crackdown, with Starlink connections active at 13 of them by spring 2026.
The companies involved point to the absence of legal incentives as the reason deeper action hasn't followed. OpenAI and Google describe robust anti-abuse programs; internet providers cite privacy constraints. A cybersecurity expert framed the core problem plainly: if there is no cost to facilitating scams, there is no reason to spend money preventing them. Outside the U.S., that calculus is shifting — the UK, EU, Australia, and Singapore have all moved toward regulations requiring companies to prevent scams or face penalties. In Washington, cooperation remains voluntary.
American technology has become the backbone of an industrial-scale fraud operation that spans continents and destroys lives with algorithmic precision. An investigation by the Associated Press and PBS Frontline found that artificial intelligence models from OpenAI and Google, combined with satellite internet service and cloud infrastructure from U.S. companies, are powering a revolution in organized crime that operates from lawless compounds in Myanmar with a reach that extends to victims in at least 17 countries.
The machinery of this fraud is both mundane and devastating. Safeer Mohammed Koorimannil, a man trafficked to a scam center in Myanmar, spent his days impersonating a 28-year-old Singaporean woman named Ella. He worked at a desk alongside dozens of others, each managing more than 100 simultaneous conversations with potential victims. Supervisors patrolled the rows with electric batons. In a single month, Koorimannil targeted approximately 50,000 people—a widowed tailor in Kurdistan, a pastry chef in Turkey, a sheep farmer in Kyrgyzstan, soldiers in Iraq, an engineer in Russia, a building painter in Germany, a port officer in Argentina, a student in Indonesia, a security guard in Poland, a dairy farmer in Georgia. He did this using software called Kongtian Intelligent Customer Acquisition, or KT, which was built on artificial intelligence models from American tech companies. The software generated responses, managed multiple conversations across dozens of languages, and tracked his performance with ruthless efficiency. When his metrics fell short, he was beaten until his body was red and swollen with lashes. "Everyone is a robot there," he told the AP from his home in southern India, speaking in his native Malayalam.
The infrastructure that enables this crime stretches across the digital supply chain in ways that have remained largely invisible to public scrutiny. While attention has focused on social media platforms where victims encounter scammers, the real machinery begins much farther upstream. American companies provide the artificial intelligence that powers the scam software, the satellite dishes that allow scammers to evade internet crackdowns, and the internet service providers that carry the traffic from Myanmar to millions of victims worldwide. The Federal Trade Commission estimates that scams cost Americans nearly $200 billion in losses in 2024. An analysis of more than 200,000 device connections at four scam compounds in Myanmar found that one in five signals was carried by a U.S.-registered company. Cogent Communications, AT&T, DigitalOcean, and Oracle all played roles in this infrastructure. Elon Musk's Starlink satellite internet service emerged as the dominant provider in Myanmar, including to known scam centers, despite public pressure from Congress and a widely publicized crackdown in the fall of 2025.
Two pieces of software illustrate how thoroughly American technology has been weaponized. KT and a similar platform called Global Social Traffic Navigation, or 007TG, were created by for-profit businesses using ChatGPT and Google's Gemini to generate automated replies, power role-play chatbots that helped scammers develop convincing personas, and embed real-time translation in over 100 languages. Blockchain analysis found that a single cryptocurrency wallet used by 007TG received $860,000 in payments between April 2024 and December 2025, with transfers from at least four cryptocurrency wallets associated with known scam networks. The scammers using these tools raked in at least $75 million. The software also tracked worker performance with devastating precision—the same metrics that got Koorimannil beaten for failing to make victims fall in love fast enough.
The human cost extends far beyond the victims who lose money. Myanmar's scam compounds have trafficked approximately 300,000 people from dozens of countries, many against their will, according to the United Nations. An Ethiopian engineer named Ebisa worked at Deko Park, a compound near the Thai border, collecting WhatsApp numbers of vulnerable men. He was beaten, shocked, detained, and forced to exercise for hours when he failed to meet impossible quotas. One day he tried to evade a beating and was caught. The security guards beat him so severely that he was blinded in one eye. Photographs documented his injuries. Speaking from a shelter for trafficking victims in Thailand in December, he said he wanted to get his eye fixed before returning home because his mother's health was fragile and he feared the sight of his injury might kill her. Months later, from his home in Ethiopia, he reported that doctors had told him it was too late. His sight could not be restored. A Nigerian man named Obinna Okeadu, also trafficked to Deko Park, never made it home. After an unsuccessful overnight shift in late October 2025, he and his roommate were called out for punishment. Back in their dorm, Okeadu began to tremble uncontrollably and collapsed. His roommate watched as he was taken away, presumably to a hospital. The next day, Okeadu's computer disappeared and his name was deleted from work chats. He was never seen again.
The victims in the United States face their own devastation. Chris Colocousis, a divorced man in his 60s from Massachusetts, was contacted on Facebook by a woman named Eliza who said she worked at a financial firm in Atlanta. She had a New York phone number. She suggested a video call, and there she was—a blond beauty with little bags under her eyes, too real not to be real. Under her guidance, Colocousis invested $400,000, money he had spent years accumulating for retirement. The investment was fraudulent. On January 25, 2025, he laid out $80,000 in cash on his table, arranged in neat rows, as instructed. A young man arrived in a Jeep with New York plates and collected the money in a plastic shopping bag. The next morning, the funds had vanished. Colocousis said he sometimes has trouble leaving his house now. "You just feel like your whole world fell apart," he told the AP. "I'm thinking about all this time that I invested into reaching a point where I could retire at a certain age—and it's just gone."
The companies involved say they lack legal and regulatory incentives to prevent the abuse. OpenAI and Google both stated they have robust programs in place to disrupt scammers, and OpenAI said it detects scams with 95% accuracy and takes down 100,000 scam accounts each month. Internet service providers emphasized that they cannot see the content their networks carry and that privacy constraints limit their ability to monitor for abuse. Starlink did not respond to detailed requests for comment. Yet the evidence suggests the companies could do more. A cybersecurity expert at Penn State University noted the fundamental problem: "If there's no disincentive to continuing this, if there's no cost to actually facilitating scamming, then why would I spend a dollar to prevent scamming?" Outside the United States, that cost is beginning to rise. The United Kingdom, the European Union, Australia, and Singapore have introduced new regulations requiring companies to prevent scams or face financial penalties. In Washington, lawmakers have asked American tech companies to cooperate on a voluntary basis. In May 2025, a Scam Center Strike Force created by the U.S. Attorney for the District of Columbia worked with Meta, SpaceX, Google, and others to disrupt more than 1.4 million social media and email accounts and seize satellite internet terminals. But the scammers have proven resilient. When Myanmar's military government demolished the high-profile KK Park compound in October 2025 and seized dozens of Starlink terminals, the scammers simply scattered and brought their technology with them. By January, at least seven devices from KK Park had migrated to a new compound 30 kilometers away near Hpakalu. Satellite imagery verified by the AP shows at least 25 new scam compounds built in Myanmar since the fall crackdown, many with white satellite dishes dotting their rooftops. Scammers from at least 13 of these new sites logged on using Starlink between March and May 2026. The infrastructure that was supposed to be disrupted had simply relocated and resumed operations.
Citas Notables
If there's no disincentive to continuing this, if there's no cost to actually facilitating scamming, then why would I spend a dollar to prevent scamming?— Sascha Meinrath, Palmer chair in telecommunications at Penn State University
You just feel like your whole world fell apart. I'm thinking about all this time that I invested into reaching a point where I could retire at a certain age—and it's just gone.— Chris Colocousis, romance scam victim from Massachusetts
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that American companies specifically are involved? Couldn't this happen with any technology?
The scale and sophistication depend entirely on access to the best tools. ChatGPT and Gemini can translate in 100+ languages and generate convincing text instantly. A scammer in Myanmar using those tools can target someone in Poland or Argentina as easily as someone next door. Without that AI, you're limited to one language, one person at a time. The industrialization only happens with American technology.
But these companies say they have safeguards. They say they detect 95% of scams. So what's the gap?
Detection happens after the fact, mostly. And even when they catch accounts, the scammers just create new ones. The real problem is that there's no cost to the companies for letting it happen. They make money either way. If you're an ISP routing traffic from Myanmar, you get paid. If you're Starlink, you get subscribers. The incentive structure is completely broken.
What about the people being trafficked? How does that connect to the technology?
The technology enables the scale of trafficking. If you need 300,000 people to run scam compounds across Myanmar, you need a system that can manage that many workers, track their output, punish underperformance. The software does all of that. It's not just fraud—it's industrial-scale human exploitation powered by algorithms.
Is there any sign this is slowing down?
No. When compounds get shut down, scammers move to new locations and bring the same equipment. Starlink usage dropped after the October crackdown, but by February it was back to being the number one provider in Myanmar. The compounds are rebuilding faster than authorities can dismantle them.
What would actually stop it?
Regulation with teeth. The UK, EU, Australia, Singapore—they're starting to impose financial penalties on companies that don't prevent abuse. That creates the disincentive that doesn't exist in America right now. But even then, scammers are adaptive. They'll find new infrastructure, new countries, new technology. This is a cat-and-mouse game that's only going to get faster.