OpenAI has banned a cluster of ChatGPT accounts used in a "task scam" originating in Cambodia. These scams involve offering highly-paid remote jobs, then requiring victims to pay their own money to access supposed earnings.

The network primarily used OpenAI's models to translate comments between Urdu and English, mimicking conversations between scammers and victims on platforms like Telegram and WhatsApp. Scammers posed as "recruiters" posting job ads and "mentors" who trained targets on how to perform tasks, often referencing luxury brands and spoofed websites to appear credible.

The workflow typically involved job postings with high salaries, recruitment via messaging apps, reassurance about legitimacy, and training on "training accounts" where "earnings" rapidly increased. Finally, victims were pressured to pay an "activation fee" to withdraw funds, with scammers making excuses to demand more money if targets hesitated.

While some targets were skeptical, evidence suggests at least some victims lost money. OpenAI has shared information about these scam networks with industry peers and authorities to combat AI-influenced adversarial behaviors and disrupt scam activities abusing their services.