OpenAI dismantles network of accounts used for fraud hubs in Cambodia

OpenAI blocked a network of ChatGPT accounts operated by criminal gangs in Cambodia, which was used for fully automated investment scams, “pig-butchering” (fake romance) and impersonating law-enforcement authorities. In a report released by the company, the model not only generated text but also built virtual identities, translated real-time communications with victims, and produced forged passports and fabricated legal documents at industrial speed.

Log analysis uncovered an additional disturbing layer: the infrastructure was used to recruit large numbers of workers under coercion, a phenomenon described as “cyber slavery”. The model disseminated job ads for chat operators promising free flights and housing, the classic lure of Southeast Asian gangs that draw young people into closed compounds and force them to scam around the clock. ChatGPT, in fact, wrote the recruitment ads that fed the system.

OpenAI handed the evidence to law-enforcement authorities and security-partner firms. The company estimates that hundreds of people fell victim, and the financial loss for individual victims reaches thousands of dollars (tens of thousands of shekel). The numbers do not include secondary damage: emptied bank accounts, erased savings, and the trauma of discovering that a virtual partner was a script running on a GPU.

The shutdown joins a series of similar actions in recent months, but it shows how low the barrier has become for running fraud at a global scale. No programmers, translators or graphic designers are needed to forge documents; a single well-crafted prompt suffices. OpenAI acknowledges that the battle is being fought on a razor-thin line: blocking the bad actors without breaking the tool for millions of legitimate users, while the pace continues to rise.