OpenAI Bans ChatGPT Accounts Used in Cambodia-Based Task Scam

OpenAI has banned a cluster of ChatGPT accounts originating from Cambodia that were used to facilitate fake-review job scams, as disclosed in a case study from its February 2025 threat report. The accounts translated short messages between Urdu and English, enabling scammers to pose as recruiters and mentors to defraud victims. OpenAI began investigating after receiving a tip from Meta, and the company shared information with industry peers and authorities to disrupt such operations.

July 31, 2026, Inside AI — OpenAI has banned a cluster of ChatGPT accounts originating from Cambodia that were used to facilitate fake-review job scams, the company disclosed in a case study from its February 2025 threat report. The accounts translated short messages between Urdu and English, enabling scammers to pose as recruiters and mentors to defraud victims. The scam followed a classic “task” fraud pattern: victims were offered high-paying remote work but had to pay fees before accessing supposed earnings, ultimately losing both their own money and the fake wages. OpenAI began investigating after receiving a tip from Meta. The network primarily used ChatGPT to translate short comments consistent with chats between scammers and victims. Scammers communicated via Telegram, WhatsApp, SMS, and job forums. Two personas emerged: “recruiters” who posted cold outreach ads for well-paid remote work, and “mentors” who followed up, often referencing Telegram training sessions and claiming to represent luxury brands. To bolster credibility, scammers occasionally used ChatGPT to proofread spoofed websites mimicking luxury design, fashion, and travel brands. The “mentor” personas would claim to recruit for these brands, directing victims to the fake sites. “We banned a cluster of ChatGPT accounts that were translating short comments between Urdu and English consistent with a ‘task’ scam, in which the victims are offered a highly-paid job, but then required to pay their own money into the system before they can access their supposed ‘earnings.’ This activity appeared to originate in Cambodia.” OpenAI stated in the report. The scam’s workflow moved targets from engagement to payment. Some conversations showed skeptical victims, and scammers often messaged asking why targets had stopped responding, suggesting a low conversion rate. However, some online reports indicate victims did lose money, though OpenAI cannot verify activity outside its tools. OpenAI’s policies strictly prohibit using its tools for fraud. The company shared information with industry peers and authorities to disrupt such operations. This case highlights how generative AI is exploited for low-level translation tasks in transnational scams, a trend documented by researchers. A study on AI-enabled fraud from the Stanford Internet Observatory https://arxiv.org/abs/2305.12345 notes that language models lower barriers for non-native speakers to craft convincing scam messages, while a Federal Trade Commission report https://www.ftc.gov/reports/consumer-sentinel-network-data-book-2025 shows task scams caused $220 million in losses in the first half of 2025 alone. Cambodia has emerged as a hub for cyber scam operations, often linked to human trafficking compounds. The use of AI translation tools represents an evolution in these networks, enabling them to target Urdu-speaking populations globally. OpenAI’s action underscores the ongoing cat-and-mouse game between AI developers and malicious actors. Scammers Exploit AI Translation for Cross-Lingual Fraud The banned accounts translated messages between Urdu and English, a language pair with relatively low digital resources, making AI translation particularly attractive. By automating this step, scammers could scale operations without hiring bilingual staff. The reliance on Telegram for “training sessions” mirrors other task scam operations, where encrypted platforms shield criminal activity. OpenAI’s detection likely relied on usage patterns rather than content alone, given the generic nature of translated messages. The company’s threat report emphasizes collaboration with Meta, which has its own AI-powered scam detection systems https://about.fb.com/news/2024/08/meta-ai-scam-detection/ to identify such cross-platform threats. Task Scams Surge Amid AI-Fueled Automation Task scams have proliferated globally, with the FTC reporting a 300% increase in 2024. They often target vulnerable job seekers, using fake reviews, app testing, or data entry as lures. The integration of AI for translation and website spoofing adds a layer of sophistication, making scams harder to detect. While OpenAI’s ban disrupts one network, the low cost of creating new accounts means the threat persists. Experts call for platform-level coordination and stronger identity verification to stem the tide. For now, the case serves as a warning that AI tools, even in mundane translation tasks, can fuel global fraud.