OpenAI says it banned ChatGPT accounts tied to a Cambodia-based scam network

OpenAI says it has disrupted a coordinated scam operation that used ChatGPT to support multiple fraud schemes, including fake investment pitches, romance scams, online gambling lures and law-enforcement impersonation. The company published the case on July 31, describing the activity as likely originating in Cambodia and likely operating in or around Poipet, a city repeatedly linked in public reporting to online scam compounds.

The disclosure is notable because it shows how generative AI can be folded into the ordinary workflow of organized fraud rather than used only for novel technical attacks. OpenAI said the network used its models to generate and translate messages sent to targets, build fake online personas, create promotional material and handle day-to-day operational tasks. The company said it banned the associated ChatGPT accounts, shared signals with industry partners and relevant authorities, and took steps intended to make it harder for the actors to return to its services.

How the operation used AI

According to OpenAI, the scammers did not appear limited to one script. Some accounts used dating personas to build trust before steering victims toward bogus cryptocurrency or spot-gold investments. Others posed as online gambling platforms offering fake winnings, or as law-enforcement agencies demanding fines for fabricated offenses. The company said users also generated images of forged documents and fake interfaces, including passports, legal notices, stock-purchase confirmations and gambling platform screens.

OpenAI framed the activity around a familiar scam sequence: first contact, emotional pressure, then extraction of money. The AI-assisted work included translation for chats on messaging platforms such as WhatsApp and Telegram, social media content for fake identities and research for dating profile material. OpenAI said the full scale of losses is unknown, though the scammers' own conversations referenced hundreds of targets and individual victims losing thousands of dollars; the company said it could not independently verify those figures.

Why it matters

The report also points to the blurred boundary between online fraud and exploitation inside some scam operations. OpenAI said some user activity suggested links to human trafficking or forced criminality, including job advertisements for roles in Poipet and records related to worker debts, salary deductions, discipline, immigration status and escape attempts. The company cautioned that it could not determine every individual's circumstances.

For defenders, the case underlines a practical challenge: shutting down AI-enabled abuse is not only about detecting malicious prompts. Scam groups can use mainstream tools for translation, persuasion, document generation and office administration. That makes account enforcement, cross-platform intelligence sharing and victim-protection work central to the response.

Sources

• openai.com

Cover photo by Sora Shimazaki on Pexels, used under the Pexels License.