Earlier this year OpenAI disrupted a coordinated scam operation that the company says very likely originated in Cambodia and relied on ChatGPT to support multiple types of fraud. The investigation began after a lead from peers at WhatsApp; OpenAI has since shared additional threat indicators with industry partners and relevant authorities.
Scope and methods of the operation
The network ran several different scams at once, including investment frauds (cryptocurrency and spot gold trading), romance scams, fake online gambling promotions, and impersonation of law enforcement. Operators routinely combined elements from different schemes — for example, using a dating persona to establish trust and then introducing fraudulent investment opportunities.
According to OpenAI, actors used the models to create and sustain fake online identities, generate and translate messages to targets, produce promotional content for fraudulent offers, and assist with daily administration. Some users also relied on ChatGPT for internal administrative tasks such as drafting announcements, translating staff messages, and documenting recruitment, immigration status, working conditions, and disciplinary matters.
Signs of trafficking and forced labor
OpenAI observed that some users produced content suggestive of human trafficking or forced labor. Examples included social-media job adverts for “chatter” roles in Poipet promising flights, accommodation, meals, visas, and work permits. The company says these findings align with extensive public reporting about organized crime groups in Southeast Asia that recruit workers with promises of legitimate employment before subjecting them to debt bondage and coercion.
Within the operation, records were kept on employee debts, salary deductions, disciplinary fines, and loan repayments. Conversations referenced immigration status, work permits, visa overstays and recruitment incentives, and in some cases mentioned apparent detention, escape attempts, and potential criminal liability for those forced to work. OpenAI notes these messages do not allow the company to independently determine the circumstances of any specific individual.
Enforcement and impact
OpenAI banned the ChatGPT accounts associated with this operation, shared relevant indicators with industry partners and authorities, and took steps to make it harder for these actors to regain access to its products and services. The full financial cost tied to the network is unknown, but OpenAI says the scammers’ own communications indicate the operation may have interacted with hundreds of targets across multiple scam types. Some internal messages referenced individual victims losing thousands of dollars, although these claims have not been independently verified.
Broader implications
OpenAI highlights two broader trends underscored by this case. First, organized scam networks are frequently diversified, running multiple fraud schemes in parallel rather than specializing in a single scam type. Second, the boundaries between online fraud, organized crime, and human trafficking are often blurred. Effective disruption therefore requires targeting not only victim-facing scam activity but also the criminal organizations that orchestrate and profit from it.
Locations and names mentioned
OpenAI identified Poipet, a city in Banteay Meanchey province, as a likely area of operation; public reporting has repeatedly linked Poipet to online scam compounds and trafficking operations.
Technical artifacts used by the scammers
The company reported that actors generated AI-created fake identities and forged documents — including passports, legal notices, stock-purchase confirmations and gambling platform interfaces — to increase the apparent legitimacy of their schemes.
