OpenAI identified and blocked a Cambodia-based scam operation that used ChatGPT to sustain romance scams, investment and betting fraud, and impersonation of authorities. The investigation began with a tip sent by colleagues on WhatsApp and led to sharing threat indicators with authorities and industry partners.
What happened
The network, very likely originating in Cambodia and linked to areas with public reports about scam compounds, used models to create and manage fake online personas, translate and generate messages aimed at victims, produce fraudulent promotional material, and support the daily logistics of the scheme.
It wasn't a single, isolated scam: criminals combined techniques and narratives. Want trust? A romantic story appears. Want speed? An 'opportunity' to invest in crypto or gold shows up. Want fear? They impersonate an authority to demand payments.
How the network operated
The operation mixed several types of deception and followed a repeatable pattern that enables large-scale fraud. We can summarize it in three clear steps:
- Ping (reach): generation and translation of conversations in WhatsApp and Telegram, creation of fake profiles on social networks and material to attract victims.
- Zing (provoke emotion): messages designed to create trust or urgency: promises of guaranteed gains, romantic language, instructions to keep things secret and offers that 'expire'.
- Sting (extract money): instructions to deposit funds, pay supposed fees or fines, and send proof of payment.
They also generated fake images and documents: passports, legal notices, receipts and screenshots of betting platforms. Some members of the group used ChatGPT for internal administrative tasks as well—ads, translations among staff, and records about debts and wages.
Indicators of possible trafficking and exploitation
Part of the activity suggested links to human trafficking and forced labor. They posted job offers for 'chatters' in specific areas promising flights, lodging and permits, and kept records about debts, salary deductions and disciplinary measures.
Some conversations mentioned arrests, escape attempts and legal threats against workers. We can't confirm every person's case, but these patterns match public reports that connect scam compounds in the region with exploitative conditions.
Important: exploiters don't only trick victims to take money; very often they also exploit the people they force to carry out the scams.
What OpenAI did and why it matters
OpenAI disabled coordinated accounts linked to this network, shared relevant indicators with partners and authorities, and took steps to make it harder for the actors to regain access to their services. There isn't a clear total of financial damage; the scammers' own conversations suggest they may have interacted with hundreds of victims and that some lost thousands of dollars.
This case reinforces two key ideas:
- Criminal networks diversify their frauds and combine tactics to maximize reach.
- There's overlap between online fraud, organized crime and trafficking; that's why the response must go beyond technical blocking and reach organizations that profit from or exploit people.
What you can learn and how to protect yourself
- Be suspicious of profiles that suddenly build trust and then ask for money or secrecy.
- Verify online job offers: listings that promise all-inclusive packages and ask you to pay for paperwork are red flags.
- Don't share sensitive documents or screenshots of payments with people you can't verify.
- Use official channels to confirm fines, prizes or legal processes before you pay.
If you think you've been a victim, contact your bank and local authorities and keep evidence of the conversations.
Technology like ChatGPT can help a lot, but in the wrong hands it increases the efficiency of existing scams. The solution combines technical measures, cooperation between companies and legal action to protect both victims and people who may be forced to work in these networks.
Original source
https://openai.com/index/disrupting-malicious-uses-of-ai-criminal-scam-operation