July 31, 2026, (Inside AI) — OpenAI has banned a cluster of ChatGPT accounts linked to a large-scale romance-baiting scam operation originating in Cambodia. The network used AI to translate and generate conversations in Japanese, Chinese, and English, targeting men over 40 through social media platforms like Facebook, X, and Instagram.
The accounts were part of a so-called "pig butchering" scheme, a type of fraud where scammers build romantic relationships online before luring victims into fake cryptocurrency or foreign exchange investments. OpenAI's investigation found that the actors used ChatGPT primarily for two tasks: generating public comments to attract targets and translating ongoing private chats.
Most public comments were posted as replies to real people's social media posts, often on content about golf or from profiles suggesting the user was a middle-aged professional, particularly in medical fields. The scammers' accounts typically featured profile pictures of young women, which OpenAI assesses were likely stolen from real models or influencers rather than AI-generated.
The operation's workflow involved moving victims from initial engagement to fraudulent investment platforms. After connecting on social media, scammers would shift conversations to encrypted messaging apps like LINE or WhatsApp, using Apple's "hide my email" function to maintain anonymity. They then introduced victims to fake investment sites, often using translated chat logs to keep the dialogue flowing.
OpenAI's report, originally published in February 2025, detailed how the network used the models to generate replies with specific personas. "The actors used our models to translate and generate short comments," the company noted. In some cases, the scammers provided extensive biographical details, including the fake persona's name, educational background, job, location, and hobbies, with a common thread being that the persona "dabbled in online investment."
The investigation uncovered evidence of many failed attempts. Some conversations ended with targets recognizing the scam, while other social media posts received no engagement. Yet, OpenAI also found references to transactions worth thousands of dollars, suggesting that in at least some cases, the fraud succeeded.
Meta had previously identified related activity on its platforms, tracing it to a newly established scam compound in Cambodia. The overlap highlights how AI tools are being weaponized in cross-platform fraud, enabling scammers to scale their operations and overcome language barriers. A 2023 study by the University of Cambridge found that AI-driven social engineering attacks increased by 300% in the past year, with pig butchering being a dominant method.
The scam's name comes from the practice of "fattening up" victims with affection before slaughtering their finances. According to the FTC's 2024 Consumer Sentinel Network Data Book, romance scams caused over $1.3 billion in losses in 2023, with many cases involving cryptocurrency. OpenAI's findings add a new layer: the industrial use of generative AI to automate the emotional manipulation at scale.
How AI Turbocharges the Pig Butchering Pipeline #
OpenAI's disruption reveals a sophisticated, multi-stage operation. The scammers first used ChatGPT to generate dozens of flirty or friendly comments on public posts, casting a wide net. Once a target responded, they moved to direct messages, where the model translated conversations in real time and even suggested replies in a specified tone of voice, such as "a flirty young woman."
This approach allowed scammers with limited language skills to engage convincingly with victims worldwide. The network's primary language was Chinese, but outputs were in Japanese, Chinese, and English. OpenAI's technical analysis showed that the model was used to generate not just translations but also culturally nuanced small talk, making the interactions feel authentic.
The scammers' toolset was diverse, including LINE, WhatsApp, X, Facebook, Instagram, and various cryptocurrency platforms. By leveraging AI, they could maintain multiple conversations simultaneously, increasing their chances of success. However, the operation was not seamless: many targets became suspicious, and some posts went ignored, indicating that even AI-enhanced scams face friction.
Industry Response and the Limits of Detection #
OpenAI stated it has shared information about the disrupted networks with industry peers and authorities. "We are dedicated to collaborating with industry peers and authorities to understand how AI is used in adversarial behaviors and to actively disrupt scam activities abusing our services," the company said. Yet, the case underscores the challenge of detecting malicious use when AI is just one link in a broader criminal chain.
Experts note that while AI companies can ban accounts, scammers quickly adapt by switching platforms or using open-source models. The Cambodia-based operation, for instance, used ChatGPT alongside other tools, making it harder to trace. Moreover, the scam's social engineering core remains human-driven, with AI serving as an amplifier rather than an originator.
In a related development, the U.S. Department of Justice announced in June 2026 a new task force focused on AI-enabled financial fraud, citing pig butchering as a priority. The initiative aims to improve coordination between tech firms and law enforcement, but critics argue that real-time intervention remains elusive given the global, decentralized nature of these crimes.