OpenAI reported in February 2025 on banning accounts that utilized AI to facilitate romance and investment scams. These accounts, which appeared to originate in Cambodia, employed AI for translating and generating conversational content. A subsequent report from OpenAI in February 2026 further detailed ongoing efforts to disrupt such malicious uses of AI, including romance scams.
Key Points
- OpenAI banned a cluster of ChatGPT accounts that translated and generated short comments in Japanese, Chinese, and English.
- This activity, consistent with romance and investment scams, appeared to originate from Cambodia.
- The actors primarily used Chinese and generated output in Japanese, Chinese, and English.
- OpenAI's models were used to generate comments for public social media posts on platforms including Facebook, X, and Instagram.
- Scammers also used the models to translate and generate short conversational comments, often providing extensive biographical details for fake personas.
- The operation frequently targeted men over 40, often in medical professions, and replied to posts about golf.
- Scammers used various tools, including LINE, WhatsApp, X, Facebook, Instagram, Apple's "hide my email" function, and cryptocurrency/foreign exchange platforms.
- OpenAI's policies strictly prohibit the use of its tools for fraud or scams.
Context
According to OpenAI, the scam activities followed a common workflow, moving targets from initial engagement to fraudulent investment. This pattern aligns with publicly available reporting on how such scams typically operate. OpenAI's February 2026 report categorized these scam workflows into three stages: the "ping" (cold contact), the "zing" (generate emotion), and the "sting" (extract money).
Why It Matters
This disruption highlights how AI models can be misused for sophisticated social engineering attacks, requiring builders and users to be aware of the evolving tactics employed by malicious actors. The detailed account of scam workflows provides insight into the practical application of AI in fraudulent schemes.
What To Do
- Note that OpenAI's policies prohibit using its tools for fraud or scams.
- Watch for guidance from OpenAI on how they identify and disrupt malicious AI use.
- Review the described scam workflows (ping, zing, sting) to understand common patterns of AI-assisted fraud.
- Be aware that scammers may use AI for translation, content generation, and persona development across multiple platforms.