OpenAI Disrupts a Poipet Scam Network That Used ChatGPT

Criminal Network Coordinated Multiple AI-Enabled Scams

OpenAI identified and blocked a coordinated network of ChatGPT accounts linked to a Cambodia-based criminal operation, likely originating in Southeast Asia and operating in Poipet.

The case was detected on August 5, 2026. The area is associated with compounds used for organized scams and human trafficking.

The network combined several schemes:

• cryptocurrency and spot-gold investment scams;
• romance scams using synthetic identities;
• fake gambling websites and services;
• law-enforcement impersonation;
• payment requests for fines or crimes that never occurred.

According to communications among the operators, the activity may have reached hundreds of victims. Total losses have not been determined; some messages referenced people who allegedly lost thousands of dollars, although OpenAI did not independently verify those claims.

ChatGPT Used to Create Profiles, Messages, and Fake Documents

The operators used ChatGPT and OpenAI’s generative models to accelerate the entire operation. The systems were used to create online identities, write and translate messages, produce advertisements, and manage internal communications.

Generated content included:

• profiles for dating websites;
• fake financial adviser identities;
• images of forged passports;
• fraudulent legal notices;
• stock-purchase confirmations;
• screenshots of gaming platforms;
• falsified proof of transfers and account information.

The strategy, referred to as “ping-zing-sting,” followed a recurring sequence: initial contact on WhatsApp or Telegram, trust-building, a request for money, and the delivery of forged documents or screenshots to make the transaction appear legitimate.

In other cases, scammers developed long-term romantic relationships before proposing fraudulent investments. Fake gaming platforms promised nonexistent bonuses and winnings, while impersonated police officers demanded urgent payments to avoid legal consequences.

Recruitment Through Fake Job Offers and Possible Forced Labor

The same infrastructure was also linked to the operation’s internal management. AI accounts drafted communications between employees and translated messages, while also recording debts, salary deductions, fines, loan repayments, and stays beyond visa limits.

Recruitment targeted users in Bangladesh and India with offers to work as “chatters” in Poipet. Packages promised:

• a base salary of $800;
• a $100 perfect-attendance bonus;
• airfare;
• food and accommodation;
• a one-year Cambodian visa;
• a work permit.

Some content was consistent with human trafficking and forced labor. Recruited individuals were allegedly stripped of their passports and forced to work in conditions resembling slavery.

Accounts Banned, but No Public Technical Indicators

OpenAI banned the network’s ChatGPT accounts and conducted the investigation in cooperation with WhatsApp, a Meta company. Telegram, social networks, and fraudulent investment and gaming services formed part of the ecosystem used by the operators.

No specific software versions, named attributions, indicators of compromise, or CVE identifiers were disclosed. No additional technical workarounds for users have been reported.

Anyone who receives WhatsApp or Telegram messages promising guaranteed investments, gaming bonuses, easy money, or urgent payments for alleged fines should end the conversation, avoid sending money or documents, and report the account to the platform. Images of passports, legal notices, and receipts do not prove authenticity.