OpenAI has identified and banned a network of accounts that used its API to generate conversations on X (formerly Twitter) and direct users to gambling websites. Dubbed operation “Bet Bot,” this network leveraged OpenAI’s models to create and manage fake personas for social media, generating bios, researching accounts to follow, analyzing posts, and drafting replies in English. These accounts, often featuring AI-generated profile pictures and sports-themed banners, frequently claimed to be based in UK cities like Manchester and Liverpool, with many created in December 2023 or June 2024.

The operation employed two main tactics: public comments on X, often replying to sports-related discussions, and direct messages (DMs). While public comments focused on sports and occasional politics to mask their true intent, the content likely intended for DMs consistently referenced gambling, sometimes including shortened URLs leading to online gambling sites. The fake personas had very low follower counts, typically in the single or low double digits, and often followed each other, artificially inflating their numbers.

While not assessed as an influence operation aimed at political manipulation, “Bet Bot” represents a modern spam network utilizing sophisticated tactics. The use of direct messaging, even with limited confirmation of clicks, suggests an ability to engage real users beyond the fake accounts' echo chambers. OpenAI places this activity in Category 2 on its Breakout Scale, indicating single-platform activity with some evidence of real-world engagement.