OpenAI reports it disrupted a Cambodia-based fraud operation centered around Poipet, where a coordinated group of accounts allegedly used ChatGPT to accelerate and expand multiple scam campaigns. The network reportedly combined romance bait, fake investment pitches, gambling lures, and law-enforcement impersonation—while using generative AI to create personas, translate and draft messages, and generate promotional content.
According to OpenAI, it banned the cluster of ChatGPT accounts involved in the Poipet scam network. The company says the group was likely originating from Southeast Asia and operating from the Poipet area, a region that has had extensive links to scam compounds and human trafficking in the past.
How the Poipet scam network used ChatGPT
OpenAI describes the operation as more than a single con. Instead, it appears the scammers adjusted their tactics to whatever narrative was most likely to work on a given target, using AI-supported workflow across fraud stages.
The banned accounts allegedly used OpenAI’s models to support the creation and maintenance of fake online identities. They also reportedly generated and translated messages sent to victims, produced promotional materials for their schemes, and handled day-to-day administrative tasks to keep the operation running.
OpenAI notes that some accounts were used for internal work such as drafting announcements, translating communications between staff, and documenting details like employee debts, salary deductions, fines, loan repayments, visa overstays, work permits, immigration status, and recruitment incentives.
Promotional ads targeting users in South Asia
A distinctive part of the operation involved promotional content for “chatter” jobs in Poipet. OpenAI says ads were created and used to target people in Bangladesh and India, offering a base salary of $800 and an additional $100 bonus for “full attendance.” The ads also highlighted benefits such as flight tickets, free accommodation, meals, a 1-year Cambodia visa, and work permits.
While the details may vary across recruitment messaging, the pattern matters: it shows how scammers blended employment bait with online engagement to reach potential victims—before shifting into romance, investment, gambling, or enforcement-style pressure tactics.
Multiple scam types run in parallel
OpenAI emphasizes that the operation illustrates a broader reality about modern scam networks: organized criminal groups often do not limit themselves to one style of fraud. They can run several scams at the same time, increasing the chance that at least one approach will succeed.
In OpenAI’s description of the Poipet scam network, that included:
- Dating-driven trust building, followed by fraudulent investment opportunities tied to cryptocurrency and spot gold trading.
- Extended romantic conversations using synthetic identities, including personas that pretended to be representatives from online gambling platforms.
- Gambling lures with fake bonuses and winnings, designed to keep targets engaged and move them toward further deposits.
- Law enforcement impersonation to create urgency, claiming victims needed to pay “fines” for alleged serious crimes.
To support these cons, OpenAI says the network created and operated fake dating profiles, fictitious investment experts, and fraudulent law-enforcement personas. The operation also reportedly generated images of forged documents, including passports, as well as legal notices, stock-purchase confirmations, and interfaces resembling gambling platform screens.
The “ping-zing-sting” scam workflow
OpenAI describes the attack chain as a three-step approach referred to as ping-zing-sting. While the terminology is specific to OpenAI’s report, the structure aligns with a common escalation pattern in online fraud.
In the first step, scammers conduct initial outreach through messaging platforms such as WhatsApp and Telegram. Next, they engage targets in a trust-building phase. Finally, they instruct victims to make payments—such as deposits, “activation fees,” or supposed fines tied to fabricated legal issues.
OpenAI also says victims were then provided with fake proof of payment, including screenshots of transfers or account information, to reinforce the illusion that money had been handled as claimed.
AI-generated content and administrative scaling
Beyond social engineering messages, the Poipet scam network reportedly used generative AI to support operational efficiency. OpenAI says some accounts produced content consistent with forced labor and human trafficking associated with organized crime groups in East Asia.
The report also points to a recruitment angle used by threat actors: promising lucrative job opportunities, then reportedly confiscating passports and forcing people into slave-like working conditions. While OpenAI notes it cannot independently verify every victim-related claim, the presence of recruitment and document-forgery themes suggests the scams could be intertwined with broader criminal exploitation.
Even where money flows are hidden behind multiple narratives, AI can help criminals move faster: drafting communications, translating messages, generating promotional materials, and keeping internal records that help a multi-person operation coordinate.
Reported victim impact and limits of verification
OpenAI states that the full scale of financial losses connected to the Poipet scam network is unknown. However, OpenAI says the scammers’ own communications suggest the operation may have interacted with hundreds of targets across multiple scam types.
The report also mentions that user conversations referenced individual victims losing thousands of dollars. At the same time, OpenAI notes it is unable to independently verify those statements.
Why this disruption matters for AI-enabled threats
This case signals how threat actors may develop AI-augmented capabilities for offensive use. OpenAI argues that modern scam networks can increase the speed and scale of their campaigns by using AI to produce content and manage workflows that would otherwise require more time and staffing.
In other words, the risk is not only that scammers can imitate individuals or generate convincing texts. The bigger concern is the ability to operate at volume: producing many personas, tailoring messages, and maintaining internal processes to support different fraud storylines at the same time.
OpenAI’s account ban is one response, but the broader lesson is that AI can be used by both defenders and attackers. As language models become more capable, criminals may find new ways to streamline social engineering, document generation, and multi-step deception.
Staying safer: signs of a multi-scheme scam
If you are dealing with strangers who pivot quickly into payments, romance, investments, or “official” demands, treat it as a red flag—especially when the communication moves through multiple platforms or includes claims that rely on urgency.
In scams like the Poipet scam network described by OpenAI, common signals include:
- Trust-building followed by payment requests (deposits, activation fees, or fines).
- Shifting narratives, such as romance leading into crypto/gold trading or gambling winnings.
- Proof documents or screenshots used as “evidence” after money is requested.
- Recruitment-style messaging promising jobs, visas, and benefits in exchange for engagement.
When in doubt, verify independently through official channels and avoid sending money or sensitive information based on chat-based claims.
Conclusion
OpenAI says it disrupted a coordinated Poipet scam network that used ChatGPT to help run a wide set of fraud schemes, from romance and investment cons to gambling lures and law-enforcement impersonation. By banning the accounts involved, OpenAI highlights how generative AI can be leveraged to create convincing personas, translate and draft messages, and scale operational tasks across multiple stages of deception.
For the public, the most practical takeaway is to recognize the pattern: scammers often blend tactics, escalate pressure, and ask for payments—while using AI-generated content to make the story feel consistent. Staying alert to these signs can reduce the chance that you become part of someone else’s workflow.
Source: https://thehackernews.com/2026/08/openai-disrupts-poipet-scam-network.html
