OpenAI has dismantled a network of ChatGPT accounts connected to a large-scale scam operation believed to be operating from Poipet, Cambodia. According to the company, the group used generative AI to support multiple fraud campaigns, including investment scams, romance fraud, fake gambling platforms, and law enforcement impersonation.
The operation is believed to have originated in Southeast Asia, with investigators pointing to Poipet, a city that has previously been associated with organized scam compounds and human trafficking.
AI Used to Power Fraud Operations
OpenAI said the scammers relied on ChatGPT for far more than writing messages. The AI tool helped create fake online identities, translate conversations into different languages, generate promotional content, and assist with routine administrative work inside the criminal organization.
Among the materials produced were social media advertisements recruiting workers for so-called “chatter” jobs in Poipet. The ads primarily targeted people in Bangladesh and India, promising salaries of around $800 per month, attendance bonuses, free flights, accommodation, meals, work permits, and one-year Cambodian visas.
Investigators believe these job offers were designed to lure victims into scam compounds operating in the region.
Internal Business Support for Criminal Networks
Beyond victim-facing content, some accounts used ChatGPT to manage internal operations.
According to OpenAI, AI assisted with drafting company announcements, translating messages between employees, tracking salary deductions, documenting debts and fines, recording visa status, and organizing recruitment activities.
This suggests AI was integrated into the group’s daily workflow rather than being limited to scam conversations.
Multiple Fraud Schemes Running Simultaneously
Rather than focusing on one type of scam, the network reportedly operated several different fraud campaigns at the same time.
Victims were approached through fake dating profiles that gradually built trust before steering conversations toward fraudulent cryptocurrency and gold investment opportunities.
Others were contacted by people posing as representatives of online gambling services, who claimed users had won prizes or bonuses that required additional payments before funds could be released.
Another part of the operation involved criminals impersonating police or government authorities. Victims received fabricated legal notices accusing them of serious crimes and were pressured into paying fake penalties to avoid arrest or prosecution.
Fake Documents and AI-Generated Identities
OpenAI said the scammers also created convincing forged materials using AI.
These included counterfeit passports, fabricated legal documents, fake investment confirmations, and fraudulent gambling platform interfaces intended to make the scams appear legitimate.
The use of AI allowed the group to rapidly produce realistic content that could be adapted for different victims and languages.
The “Ping-Zing-Sting” Strategy
Researchers described the group’s approach as a three-stage process known as “ping-zing-sting.”
The scam typically began with an initial message sent through platforms such as WhatsApp or Telegram. Once contact was established, scammers spent time building trust before convincing victims to transfer money through investments, activation fees, or fabricated legal fines.
To reinforce the deception, victims were shown fake screenshots of successful transactions or account balances designed to create the illusion that payments had been processed.
Links to Human Trafficking Concerns
Some of the recovered conversations also pointed to possible links with human trafficking.
OpenAI noted that certain recruitment messages resembled tactics used by organized crime groups that lure people abroad with promises of high-paying jobs, only to confiscate their passports and force them to participate in online scam operations under abusive conditions.
These practices have been widely documented across parts of Southeast Asia in recent years.
Investigation Conducted With WhatsApp
OpenAI said the investigation was carried out in collaboration with WhatsApp, which is owned by Meta.
The company emphasized that modern criminal organizations often operate several different scam models simultaneously, switching tactics depending on what is most likely to deceive potential victims.
Growing Role of AI in Cybercrime
While OpenAI removed the accounts involved, the incident highlights a growing concern within the cybersecurity community.
Generative AI is increasingly being used by cybercriminals to automate phishing campaigns, create convincing fake identities, translate messages across languages, and scale fraud operations far more efficiently than traditional methods.
Although AI platforms include safeguards against abuse, security experts warn that attackers continue searching for new ways to exploit these technologies, making ongoing monitoring and rapid detection increasingly important.