OpenAI has disabled a number of ChatGPT accounts allegedly operated by a Cambodian criminal network that used the artificial intelligence chatbot to carry out fraudulent cryptocurrency investment schemes. This action comes as digital asset-related scams proliferate, prompting exchanges, token issuers, and AI companies to bolster their security measures to protect potential victims.
AI-powered scams on the rise
Southeast Asia, particularly Cambodia, has emerged as a hotspot for so-called “pig-butchering” scams, where fraudsters use fake online relationships or investment offers to extract crypto payments from victims. Recent scams increasingly deploy AI-generated content to enhance credibility and scale up operations.
OpenAI began its investigation after receiving a tip-off from WhatsApp. The traced accounts, believed to be based in the border city of Poipet, were connected to established scam operations already under international scrutiny.
The company reported that the network used ChatGPT to create counterfeit identities, translate communications into multiple languages, and craft persuasive messages aimed at earning the trust of unsuspecting individuals. In addition, the AI tool generated polished marketing materials and documents to lend legitimacy to fraudulent investment opportunities.
At the core of the scheme was the use of cryptocurrencies. Victims were initially engaged in social interactions before being directed to fraudulent platforms for cryptocurrency or spot-gold trading. The network leveraged ChatGPT to produce fake passports, legal notices, stock confirmations, and fabricated trading dashboards to further mislead victims.
OpenAI detailed that the fraudulent activity typically unfolded in three stages: victims were first convinced to send funds in cryptocurrency, then pressured to provide additional payments under the pretense of activation charges, taxes, or regulatory penalties.
Links to human trafficking operations
OpenAI’s examination uncovered evidence of labor trafficking embedded within the scam operation. Certain entities used ChatGPT to design appealing advertisements for recruiting “chatters” in Poipet, offering benefits such as free transportation, housing, meals, and visa assistance.
Further, internal communications outlined efforts to manage and penalize workers, tracking employee debts, salary deductions, fines, and even details on immigration issues or attempted escapes.
These findings are consistent with research by Elliptic, a blockchain analytics firm, which identified that many perpetrators behind romance and investment scams in the region are themselves victims of human trafficking, forced into organized criminal labor.
Prior to shutting down the accounts, OpenAI shared its discoveries with industry partners and law enforcement. The company estimates that hundreds of individuals have fallen victim to the scam, with cumulative losses reaching thousands of dollars.
The incident highlights a broader trend in cybersecurity, as organized crime networks increasingly pair advanced AI tools with cryptocurrency payment channels. Curbing illegal use of such technologies has become as critical as targeting illicit financial activity directly.
Mini dictionary: Elliptic is a blockchain analytics company that specializes in detecting and investigating crypto-enabled financial crime for businesses, financial institutions, and law enforcement.
Record illicit crypto flows and new chokepoints
TRM Labs, a blockchain intelligence firm, reported that illicit cryptocurrency flows are projected to reach $158 billion by 2025, marking a 145% increase over the previous year and the highest total in five years. Although the share of illicit activity within total on-chain volume dropped from 1.3% in 2024 to 1.2% in 2025, stablecoins accounted for about 84% of the detected fraudulent inflows, making them the primary payment channel for criminal operations.
| Year | Illicit Crypto Flows | Share of On-Chain Volume |
|---|---|---|
| 2024 | $64.6 billion | 1.3% |
| 2025 | $158 billion | 1.2% |
The removal by OpenAI addresses only one network but illustrates how AI platforms have become pivotal in both enabling and fighting against increasingly sophisticated scam operations.
AI’s expanding role in fraud and prevention
Recent research involving international universities revealed that AI-powered chatbots can outperform human fraudsters in persuading victims to follow scam instructions. One experiment showed that 46% of participants trusted and acted on advice from an AI agent, compared to 18% who engaged with a human scammer. The results highlighted that these participants considered the AI far more trustworthy than its human counterpart.
The study included interviews with 145 individuals previously involved in scam operations, many of whom were survivors of trafficking in Cambodia, Myanmar, and Laos. The researchers found that AI’s primary benefit to criminal networks lies in automating the crucial trust-building stage. Blocking access to sophisticated AI models, therefore, has the potential to disrupt fraudulent schemes before they secure victims’ funds.
OpenAI’s internal analysis concluded that members of the identified network reached out to hundreds of victims, who collectively lost thousands of dollars as a result of the scam.
