OpenAI Shuts Down Cambodian Fraud Network That Exploited ChatGPT

In August 2026, OpenAI publicly disclosed that it had shut down the activity of a criminal group based in Cambodia. Its members used ChatGPT as a tool to automate and scale four types of fraud at once: investment schemes, romance fraud, illegal gambling and impersonation.

How the scammers used ChatGPT

The criminal network used ChatGPT to automate schemes aimed at victims in different countries around the world. According to a post on the OpenAI blog, the operation covered four key areas:

• Investment fraud — ChatGPT generated convincing scripts for talking to victims, along with "backstories" for nonexistent brokers and traders
• Romance fraud — creating fake conversations on behalf of nonexistent people to win trust and then defraud victims financially
• Illegal gambling — drawing users to prohibited gambling platforms
• Impersonation — posing as real people or legitimate organizations

The combination of the first two types is the classic "pig butchering" scheme: the scammer builds a long "romantic" or "friendly" relationship with the victim through messaging apps and social networks, then urges them to invest in a fake cryptocurrency or investment platform. According to the US Federal Bureau of Investigation (FBI), Americans lost more than 4.5 billion dollars to this scheme in 2024, making it one of the most profitable forms of modern financial fraud.

Generative AI has substantially lowered the skill threshold for such operations: ChatGPT handles the language work, letting operators hold conversations with victims in English, Chinese, Japanese or Korean without actually speaking those languages. A single operator can use AI to sustain dozens of conversations at once.

Why Cambodia

In recent years, Cambodia has become one of the largest regional hubs of organized online fraud in Southeast Asia, alongside Myanmar and Laos. According to the United Nations Office on Drugs and Crime (UNODC), "scam farms" operate in special economic zones in these countries: closed compounds where tens of thousands of people work as operators of criminal schemes under coercion, often after falling victim to labor trafficking themselves.

OpenAI identified the accounts involved in the operation that violated its acceptable-use policy and blocked them. This is not the company's first public report on disrupting such schemes: since 2023, OpenAI has regularly published data on detected abuses of ChatGPT, ranging from state-run influence campaigns to criminal fraud schemes. Each report describes the type of scheme and the rules violated, without disclosing technical details of the detection system.

According to the official

OpenAI blog, the company detected the criminal operation and took action to stop the violations of its services' usage policy.

What this means

The disruption of the Cambodian operation is a clear example of how large language models have become an infrastructure element of organized crime, on a par with anonymous messengers and payment systems. Notably, the group combined four types of schemes in a single operation, which points to a high level of organization: AI tools are being used not by random scammers but by structured criminal organizations. AI developers have to build abuse-monitoring systems at a scale comparable to the antivirus industry, and that demand will only grow.