OpenAI shuts Cambodia-linked ChatGPT accounts over crypto scams

OpenAI has reportedly terminated a series of ChatGPT accounts that are linked to a Cambodian criminal organization that used the chatbot to commit fraud in cryptocurrency investment. It presents another case where AI and crypto have been misused by organized criminals.

This comes in when digital assets-linked scams are on the rise, which is now urging exchanges, token developers, and artificial intelligence companies to take action to prevent losses to victims. “Pig-butchering” scams are already widespread in Southeast Asia, where fraudsters swindle crypto payments with either dating or investment schemes while using AI-generated content.

How the operators wired ChatGPT into the con

According to OpenAI’s report, the investigation began after receiving a tip-off from WhatsApp, and they have since traced the activity back to accounts alleged to be based in Poipet, a Cambodian border city already known to international investigators for its connection to large-scale scams.

According to the firm’s statement, the operators exploited ChatGPT by creating fake identities, translating talks into different languages, and composing messages to win people’s confidence. Furthermore, the chatbot produced marketing materials for the fake investment schemes.

At the heart of the plot was cryptocurrency. Victims were lured into love or friendship conversations, then redirected to fake cryptocurrency and spot-gold trading platforms. To further add to the illusion, the perpetrators employed ChatGPT to fabricate fake passports, legal notices, stock purchase confirmations, and phony crypto trading dashboards.

The AI firm indicated that the fraud occurred in three stages, which allowed the criminals to trick victims into sending them cryptocurrency first and then request more money through some supposed activation fees, taxes, or regulatory fines.

The trafficking layer sitting behind the accounts

OpenAI said its investigation also uncovered evidence of labor trafficking.

Some companies have taken advantage of ChatGPT to design ads for recruiting “chatters” in Poipet. The ads include attractive offerings like free tickets, lodging, food, and help with visas. Other discussions have to do with making spreadsheets to keep track of employees’ debts, money taken from their salaries, fines for inappropriate behavior, immigration status, detention, or attempted escape.

The results bear a resemblance to the study conducted by Elliptic, whereby most of the actors behind romance and investment scams happen to be victims of human trafficking made to work in organized crime operations throughout the Southeast Asian region.

Before shutting down the accounts, OpenAI provided information from the investigation to industry stakeholders and law enforcement. The company surmises, based on internal communication, that the network has reached out to hundreds of victims and that their losses may total thousands of dollars.

This case illustrates a larger trend regarding cybersecurity. As organized crime syndicates can utilize generative artificial intelligence and cryptocurrency payment systems, preventing criminals from using advanced AI tools might become as key as blocking illegal wallets or shutting down money laundering operations. 

Illicit crypto flows hit a record just as the ban dropped

Given the scale of crypto crime, OpenAI’s move is huge. According to TRM Labs’ Crypto Crime Report released on January 28, illicit cryptocurrency flows were estimated to amount to $158 billion in 2025. This was an increase of nearly 145% from the previous year and is the largest figure in five years. While the share of illicit activity within on-chain volume dropped from 1.3% in 2024 to 1.2% in 2025. This indicates that crypto has been adopted more widely. 

The report also indicates that stablecoins are responsible for around 84% of verified fraud money inflow, confirming their being the number one method of payment for criminals engaged in organized scams. However, OpenAI’s shutdown is aimed at a single operation done within a criminal infrastructure.

Why AI companies have become a chokepoint for scam networks

OpenAI’s efforts are huge because generative AI can already carry out the most labor-intensive part of scams – convincing victims to trust scammers.

A study by researchers from Amrita Vishwa Vidyapeetham, Ca’ Foscari University of Venice, University of Melbourne, and Ben-Gurion University of the Negev indicated how an AI-enabled Claude agent managed to outperform a human scammer in the process of making a pig-butchering scam. It revealed that 46% of the participants used the software suggested by the AI after one week of messages, while only 18% of people used it with the human operator. Moreover, the participants concluded that the AI was much more reliable.

The researchers conducted interviews with 145 people who were previously involved in scamming operations, including survivors of trafficking operations in different facilities in Cambodia, Myanmar, and Laos. According to the findings, one of the functions of using AI in scam operations is its ability to automate the trust-building stage in the scams, and blocking access to sophisticated models can lead to the disruption of the operation process even before the victims have been influenced to give up their money.

The smartest crypto minds already read our newsletter. Want in? Join them.