OpenAI has banned a coordinated set of ChatGPT accounts tied to a scam operation based in Cambodia[1]. What stands out is that the group ran several schemes at once, covering investment fraud, romance scams, bogus online gambling offers, and impersonation of law enforcement. The investigation started with a tip from WhatsApp, and the indicators OpenAI uncovered have since been shared with industry partners and relevant authorities.

A Tip From WhatsApp Started the Investigation

Earlier this year, OpenAI disrupted a scam operation based in Cambodia. The company began looking into the activity after a lead from its peers at WhatsApp, and has since passed additional threat signals to industry partners and relevant authorities[1].

The accounts that were banned formed a coordinated network. OpenAI says the cluster very likely originated in Cambodia and was likely operating in or around Poipet, a city in Banteay Meanchey province that public reporting has repeatedly linked to online scam compounds and trafficking operations[1].

The network used ChatGPT in four ways: creating and running fake online personas, generating and translating messages sent to targets, producing promotional content for the fraudulent schemes, and handling day-to-day operations[1].

Investment, Romance, Gambling, and Police Impersonation Ran in Parallel

The most notable feature of this case is that it was never confined to a single type of scam. OpenAI notes that organized criminal groups rarely restrict themselves to one scheme. Instead they opportunistically pick whichever narratives, personas, and tactics they expect to work best on a given target[1].

In practice, the network ran multiple scams at the same time and often blended elements of each. Operators used dating personas to build trust before introducing fraudulent investment opportunities involving cryptocurrencies and spot gold trading. Other users held lengthy romantic conversations under fictitious identities, posed as representatives of online gambling platforms offering fake bonuses and winnings, or impersonated law enforcement agencies and told targets they had to pay fines for serious criminal offenses[1].

The narratives varied, but the underlying pattern of deception did not. The operators built out fake dating profiles, fictitious investment experts, and fraudulent law enforcement personas, and they went as far as generating images of forged documents, including passports, legal notices, stock-purchase confirmations, and gambling platform interfaces[1]. Even the paperwork was handed off to generative AI.

The Ping, the Zing, and the Sting

OpenAI describes the mechanics of this kind of fraud using a three-stage pattern it has referenced in previous reports: the ping (outreach), the zing (generating emotion), and the sting (extracting money)[1].

In the ping stage, ChatGPT was used to translate and generate conversations with targets on messaging platforms such as WhatsApp and Telegram. The same applies to creating social media content and researching material for the dating profiles that backed up the fake personas[1].

The zing stage centers on emotional pressure and trust-building. Observed examples include promises of guaranteed returns and risk-free investments, romantic language, instructions to keep the conversation secret, and urgent requests to act before a fictional bonus expired[1].

Then comes the sting: instructing victims to make deposits to unlock supposed rewards, pay activation fees, or settle fictitious fines, and then to send screenshots of transfers or account details as proof of payment[1].

Signs of Human Trafficking

Separately from the scams themselves, some users in the network generated content suggesting links to human trafficking or forced labor. One example was social media advertising for chatter jobs in Poipet that promised flights, accommodation, meals, visas, and work permits[1].

Other activity looked like internal workforce administration. Users maintained records of employee debts, salary deductions, disciplinary fines, and loan repayments, and translated discussions about immigration status, work permits, visa overstays, and recruitment incentives. Some conversations referenced apparent detention, escape attempts, and potential criminal liability for people who had been trafficked and forced to work in scam operations[1].

OpenAI says it cannot independently determine the circumstances of any particular individual, but notes that these observations are consistent with extensive public reporting on organized crime groups in Southeast Asia, which recruit workers with promises of legitimate employment before trapping them in systems of debt bondage and coercion. The reminder that the people running these scams can themselves be victims of exploitation is worth sitting with[1].

Scale of the Damage and Two Emerging Trends

OpenAI banned the ChatGPT accounts associated with the operation, shared relevant indicators with industry partners and relevant authorities, and took steps to make it harder for these actors to regain access to its products and services[1].

The full scale of the financial losses is unknown. Based on the scammers' own communications, however, the operation may have interacted with hundreds of targets across multiple scam types. Some conversations referenced individual victims losing thousands of USD (roughly several hundred thousand yen at 159 JPY to the dollar), though OpenAI says it cannot independently verify those claims[1].

※1 USD = 159 JPY (as of July 31, 2026)

The company draws two trends from the case. First, organized scam networks can be highly diversified, running several fraud schemes simultaneously rather than sticking to one. Second, the boundaries between online fraud, organized crime, and human trafficking are often blurred. That leads to a clear conclusion: effective disruption has to target not just the victim-facing scam activity, but the criminal organizations that orchestrate and profit from it[1].

Discussions of generative AI misuse have tended to focus on disinformation and more convincing phishing email. What this report shows is a stage earlier than that, with AI embedded in the business operations of a scam compound, from workforce administration to job advertising. Banning accounts treats the symptom, so the real question is how well platform-side detection can be meshed with information sharing across companies and authorities.

Summary

OpenAI has banned the ChatGPT accounts of a scam network believed to have operated in and around Poipet, Cambodia. The group ran investment, romance, gambling, and law enforcement impersonation schemes at the same time, using ChatGPT for everything from running fake personas and generating forged documents to managing its own workforce. Traces suggesting human trafficking were also found, underlining how closely online fraud and organized crime sit together.

出典[1] https://openai.com/index/disrupting-malicious-uses-of-ai-criminal-scam-operation/