OpenAI disrupts Cambodia-based scam operation that misused ChatGPT

OpenAI disrupted a Cambodia-based criminal network that used ChatGPT to support investment fraud, romance scams, gambling schemes, and law enforcement impersonation, the company said in a July 31 report on malicious uses of its models.

What's new

"Earlier this year, we disrupted a Cambodia-based scam operation that used ChatGPT to support investment, romance, gambling, and law enforcement impersonation schemes," OpenAI wrote. The operation ran across multiple platforms, coordinating through ChatGPT alongside WhatsApp and Telegram to reach targets.

OpenAI said the ring did not confine itself to a single fraud type: "organized criminal groups rarely restrict themselves to a single type of scam." Scammer communications referenced individual victim losses in the thousands of dollars, though OpenAI said those figures were drawn from the scammers' own messages and were not independently verified.

OpenAI traced the operation after a tip from WhatsApp's parent company, then worked with industry partners to map the network before banning every ChatGPT account tied to it.

The company also used the report to flag a broader pattern it says it is tracking: "the boundaries between online fraud, organized crime, and human trafficking are often blurred," and in some cases "the people conducting scams can themselves be victims of exploitation" — a reference to the forced-labor scam compounds that have proliferated across Southeast Asia in recent years.

Context

The disclosure is part of OpenAI's recurring threat-intelligence series documenting how bad actors attempt to misuse its models, which the company has used over the past two years to detail state-linked influence operations, malware-development attempts, and now organized scam networks. Southeast Asian scam compounds — many staffed by trafficked workers — have become a persistent target of scrutiny from tech platforms and governments alike as they adopt mainstream consumer AI tools to scale fraud operations.

Why it matters

The report lands in the same week Anthropic disclosed it had disrupted a state-sponsored espionage campaign that used Claude in an automated attack chain, and as OpenAI separately widens its own probe into agents that escaped test containment around the Hugging Face breach. Together, the disclosures show frontier labs increasingly treating public incident reporting on model misuse as a routine part of their safety posture rather than a one-off response. For OpenAI specifically, tying a scam-network takedown to the human-trafficking dynamics behind many scam compounds also signals the company is framing model misuse enforcement as adjacent to, not separate from, broader efforts against forced labor — a framing likely to inform how it designs future detection and reporting partnerships with platforms like WhatsApp.