The tools Silicon Valley built to automate customer service are now automating heartbreak at industrial scale.
The Summary
- A trafficked worker in Myanmar used AI-powered software from American tech companies to romance-scam 50,000 victims across 17 countries in one month, managing over 100 simultaneous conversations per shift.
- Scam operations have industrialized using U.S. tech company AI models, enabling unprecedented speed and global reach that watchdogs say could be stopped if companies chose to act.
- The targets ranged from a widowed Kurdish tailor to a German painter to a Georgian dairy farmer, showing how AI has democratized vulnerability worldwide.
The Signal
Safeer Mohammed Koorimannil had four days to make each mark fall in love. Working under armed supervision in a Myanmar compound, he posed as "Ella," a 28-year-old Singaporean woman, across dozens of profiles simultaneously. This wasn't skilled social engineering. It was assembly-line fraud, powered by AI systems built by American tech companies and repurposed for industrial-scale deception.
The math is staggering. One operator, one month, 50,000 targets. That's not humanly possible without automation. The software let Koorimannil maintain coherent romantic conversations with over 100 people at once, each believing they'd found genuine connection. His victims weren't clustered in one region or demographic. They were everywhere: soldiers in Iraq, engineers in Russia, students in Indonesia, pastry chefs in Turkey, sheep farmers in Kyrgyzstan.
"Everyone is a robot there."
An AP and "FRONTLINE" investigation revealed that U.S. tech companies' AI models are central to what amounts to the industrialization and globalization of fraud. This isn't about criminals finding clever workarounds. This is about infrastructure. The same large language models that write marketing emails and customer service responses now write love letters at scale, learning what emotional triggers work on which demographics, optimizing for maximum extraction.
What makes this particularly dark is the human layer. These aren't willing scammers. Many operators are trafficked workers, held in compounds with electric batons, forced to exploit strangers using tools that make their quotas achievable. The AI doesn't just enable the scam. It enables the entire labor model behind it, lowering the skill floor so dramatically that coerced workers can hit targets that would be impossible through human effort alone.
Key factors enabling this:
- AI models that handle context across dozens of simultaneous conversations
- Natural language generation good enough to pass as human emotional connection
- Software designed for customer engagement, now weaponized for romance fraud
- No effective safeguards preventing bulk account creation or pattern detection
The watchdog assessment is blunt: these companies have the technical capacity to do more. Rate limiting, behavioral analysis, coordination across platforms, all technically feasible. The question isn't capability. It's priority. Fraud detection eats into growth metrics. Friction reduces engagement. And so the infrastructure remains open for abuse, built for scale and accidentally perfect for crime.
This is what happens when you optimize for engagement without considering misuse at the margins. The margins just became the majority use case in certain geographies. Web2 gave us the read-write internet. These scammers got read-write-deceive, industrialized, with American technology doing the heavy lifting.
The Implication
If your company builds tools for automated communication at scale, you're building dual-use technology whether you admit it or not. The romance scam economy is a preview of what's coming: AI that doesn't just assist human fraud, but makes human fraud operators obsolete except as button-pushers. Watch for regulatory pressure in 2025 as governments realize their citizens are being targeted by AI systems trained by their own tech companies. The safeguards will come eventually. The question is how many billions get extracted first.
For everyone else: if someone you've never met in person wants money, it's not romance. It's a job. And the person on the other end might be a trafficked worker in Myanmar with a quota to hit and AI doing most of the talking.