AINEWS
The US Tech Stack Behind Myanmar's Scam Factories
A trafficked worker's smuggled notebooks expose the American AI models and internet infrastructure powering industrial fraud across 17 countries.
A trafficked worker's smuggled notebooks expose the American AI models and internet infrastructure powering industrial fraud across 17 countries.
Koorimannil was given four days. Four days to make a stranger fall in love with a persona called Ella, a fabricated 28-year-old Singaporean woman he operated from a scam compound on the Myanmar border. He ran dozens of profiles at once, chatting with more than 100 people simultaneously while supervisors carrying electric batons walked the rows of desks. In roughly one month he targeted about 50,000 people across at least 17 countries, according to records he smuggled out of the compound and shared with The Associated Press and FRONTLINE (AP/FRONTLINE investigation).
His notebooks, corroborated by the AP/FRONTLINE reporting, name the people on the other end of those chats: 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, and a dairy farmer in Georgia (syndicated AP wire in the Durango Herald). The four-day script was designed to compress courtship into a production cycle. Koorimannil described the operation as machine-like. "Everyone is a robot there," he told AP from his home in southern India in his native Malayalam (AP/FRONTLINE investigation).
The force multiplier is American. The compound's software for persona generation, multilingual outreach, and scripted escalation was built on AI models from US tech companies, the same category of large language models now sold to legitimate enterprises worldwide. According to the AP/FRONTLINE investigation, much of the cross-border internet traffic into Myanmar's border regions runs over American internet providers, and satellite links have been used to evade crackdowns, putting the US digital supply chain, not just the scripts, inside the pipeline (AP/FRONTLINE investigation). The result is that human trafficking inside the compound becomes a high-throughput fraud operation, with each trafficked worker able to run dozens of simultaneous "relationships" that would have required an entire small business a decade ago.
The pipeline's economics matter because they decide who can intervene. As the AP/FRONTLINE investigation reports it, watchdogs argue that the same US AI vendors whose models enable the targeting have the technical capacity to detect and throttle industrial-scale persona generation but lack the legal mandate, regulatory pressure, or commercial incentive to do so. The firms sell general-purpose models and have written usage policies against fraud; the watchdogs' point is that the policies have not been matched by enforcement that keeps pace with deployment (AP/FRONTLINE investigation).
Koorimannil's testimony is one worker's account, not an industry census. The 50,000-victims-in-a-month and 17-country figures are his claimed work product per the records he smuggled out, and the operational details, from the electric batons to the four-day rotation, rest on him and those documents. What his notebooks do is make the supply chain legible. The gap between what US-built AI tooling can technically do and what its providers are structurally incentivized to prevent is now the upstream condition for industrial romance fraud, and the next test is whether any of those providers treats compound-scale targeting as a measurable product-safety problem rather than a terms-of-service footnote.