Inside the Cambodia-Based Scam Network That Used ChatGPT to Run Investment Fraud, Romance Cons, and Impersonation at Scale

The four scam types the network ran

OpenAI identified that the same coordinated network of accounts was used to support at least four distinct fraud types :

1. Investment fraud (crypto and spot gold). Operators used dating personas to build trust with targets before introducing fake investment opportunities in cryptocurrencies and spot gold trading. They used ChatGPT to generate fake stock-purchase confirmations and trading interfaces that appeared legitimate.

2. Romance scams. Users engaged in lengthy romantic conversations with targets using fictitious identities, a technique commonly known as "pig butchering" where scammers groom victims over weeks or months before introducing a financial ask.

3. Gambling scheme fraud. Scammers posed as representatives of online gambling platforms, offering fake bonuses and winnings to lure victims into depositing money that was never recoverable.

4. Law enforcement impersonation. Operators impersonated law enforcement agencies to tell targets they needed to pay fines for serious criminal offenses. ChatGPT was used to generate forged legal notices, passports, and other official-looking documents.

Evidence linking the operation to human trafficking and forced labor

Perhaps the most disturbing aspect of OpenAI's findings was evidence that the people running these scams may themselves have been victims of exploitation. OpenAI's report detailed several indicators consistent with human trafficking and forced labor, as documented by Amnesty International, the United Nations, and other organizations :

• AI-generated job advertisements for trafficked workers. Users created social-media ads for "chatter" jobs in Poipet that promised flights, accommodation, meals, visas, and work permits — a known recruitment tactic used by trafficking rings to lure foreign nationals into scam compounds .

• Records of employee debt and coercion. Users maintained spreadsheets tracking employee debts, salary deductions, disciplinary fines, and loan repayments. Internal communications translated discussions about immigration status, work permits, visa overstays, and recruitment incentives .

• References to detention and escape. Some conversations referenced apparent detention, escape attempts, and the potential criminal liability facing trafficked individuals forced to work in scam compounds .

• Consistency with extensive public reporting. OpenAI noted these indicators are consistent with reporting from the Wall Street Journal, Amnesty International, Reuters, and the Associated Press describing organized crime groups in Southeast Asia that recruit workers with false promises of legitimate employment, then trap them in debt bondage and coercion .

The company stated plainly :

"While we cannot independently determine the circumstances of every individual involved, the activity serves as a reminder that the people conducting scams can themselves be victims of exploitation."

How OpenAI disrupted the operation

OpenAI's response had three main components :

1. Account bans and access denial. The company banned the coordinated network of ChatGPT accounts and implemented measures to make it harder for the actors to regain access to OpenAI's products and services.

2. Intelligence sharing. OpenAI shared relevant threat indicators with industry partners and relevant authorities, following the initial lead from WhatsApp.

3. Public disclosure. The company published a detailed threat disruption report so that other platforms, security researchers, and law enforcement agencies could benefit from the findings.

Scale and financial impact

The full scale of financial losses is unknown, but OpenAI's report noted that based on the scammers' own internal communications, the operation may have interacted with hundreds of targets across multiple scam types, with individual victims losing thousands of dollars .

Broader context: Cambodia's scam economy

This operation is not an isolated incident. According to a 2024 United Nations report, an estimated 100,000 to 150,000 people, many of them foreign nationals deceived by fraudulent job offers, are believed to have been trafficked and held in Cambodian compounds where they are forced to engage in online scam operations . Amnesty International's 2025 report documented more than 50 scamming compounds across Cambodia as sites of human trafficking, slavery, and forced labor .

OpenAI had previously disrupted related activity. In February 2025, the company banned a cluster of ChatGPT accounts whose activity appeared connected to an operation originating in Cambodia, using AI to translate and generate comments for a romance baiting network across social media and communication platforms including X, Facebook, Instagram, and LINE . Meta's subsequent investigation indicated the malicious activity appeared to originate from a "newly stood up scam compound in Cambodia" .

What this means for AI safety and platform security

The Cambodia network case study illustrates a growing challenge for AI companies: the same language models that power productivity, creativity, and learning can also be weaponized by sophisticated criminal networks. OpenAI's approach — combining account bans, intelligence sharing, and public transparency — represents one model for how AI providers can fight back without relying solely on reactive moderation.

The case also highlights a deeper problem: account bans alone cannot dismantle physical operations or rescue trafficking victims. As one industry analysis noted, OpenAI disrupted an operational support layer, not the physical organization behind the scams . The people trapped in those compounds remain vulnerable, and the structural issues that enable these networks demand coordinated responses from governments, law enforcement, and civil society.