OpenAI said on February 1, 2026 that it had banned accounts that used AI to support romance scam workflows, citing examples of outreach, translation, victim engagement and investment-fraud lures. The company published a case study as part of a report on disrupting malicious uses of AI.
Three-stage scam pattern
According to OpenAI, these scams often follow a three-step pattern the company labels “ping,” “zing,” and “sting.” The “ping” is a cold contact designed to attract a potential target’s attention—examples include replies to social media posts, cold-call SMS, fake recruitment messages or social media ads.
The “zing” step aims to generate emotion, such as attraction in romance scams or excitement about a supposed investment. The “sting” is the phase that attempts to extract money, using pretexts like a loved one’s business, a financial emergency, or a supposed investment opportunity.
AI role and disrupted cases
OpenAI said threat actors used ChatGPT to craft messages that could seem more engaging and less obviously non-native than traditional scam texts. The company referenced a 2025 disruption of a newly established operation in Cambodia that used ChatGPT to generate romance-scam messages, some spread on social media.
Other examples cited include a “pig butchering” network that targeted American men in their 40s who posted about golf, and an operation labeled “Wrong Number” that used cold-call SMS with promises of implausibly high returns. OpenAI also noted a romance scam that used social media ads to target young men in Indonesia.
OpenAI emphasized that distribution methods vary—SMS, encrypted messaging apps, social media posts and ads—and that some operations use scattershot distribution while others attempt more precise targeting. The company assessed that the chosen distribution method affects a scam’s ability to reach and exploit targets, independent of the degree of AI use.
The case study was published as part of OpenAI’s February 2026 report on disrupting malicious uses of AI and follows prior disclosures made in June 2025 and February 2025 about other scam networks.
Original source: OpenAI News