AI-Powered Fake Dating Apps Charged Users to Chat With Bots, Anthropic Investigation Finds
Anthropic found 28 dating apps using 4,700+ AI personas to charge users for chats with supposed singles, reaching at least 25,000 people and generating 2.36 million messages in April 2026. The scheme surfaced when a five-day-old prepaid account sent 100,000+ daily requests to Claude.
A network of fraudulent dating apps used AI-generated personas to deceive users into paying real money to chat with bots they believed were real people, according to findings published by AI company Anthropic in September 2026. The operation involved roughly 28 apps, more than 4,700 fabricated AI personas, and at least 25,000 real users over a two-week period in April 2026 alone — generating approximately 2.36 million messages.
Anthropic first detected the scheme after spotting a five-day-old prepaid account sending over 100,000 API requests per day to its Claude model. Researcher Chris Cronbaugh presented the findings publicly at a cybersecurity conference called Sleuthcon in June 2026, in a talk titled “Swipe Right, Pay Up: Industrial-Scale AI Catfishing.” The apps — including Dora, Romi, Doni, GraceChat, Jovia, Luma, Kira, and Nalo — were presented as legitimate dating platforms promising “real connection” with nearby singles.
The scam worked by mixing a small number of real gig workers — who passed video liveness checks and selected from pregenerated reply options — with fully autonomous AI personas that kept conversations going around the clock. Users paid for in-app “coins” or “gems” to continue chatting, believing they were talking to real women. The majority of interactions involved no human at all. Backend systems fabricated likes, profile visits, and pre-recorded video calls to maintain the illusion.
Security researcher Matthew “Zigula” Gore-Kormanik, who analyzed the apps independently, found an internal operations manual — written in Chinese and accidentally shipped inside the Doni app — that detailed how gig workers were monitored, ranked, and paid. Anthropic attributes the operation to a China-based actor, citing Chinese-language materials, Chinese cloud infrastructure including Tencent Cloud, and source code hosted on Chinese platform Gitee. The apps did not function as scams inside China or in Asia.
Most apps were removed from the Apple App Store and Google Play Store between August and early September 2026, though Kira remained live on Google Play as of September 16th. Anthropic said it shared findings directly with Apple and Google. Tate Jarrow, CEO of anti-scam app Jacana and a former U.S. Secret Service cybercrime investigator, noted that sophisticated scammers now operate like businesses, using AI to cut costs and scale operations — making cross-industry collaboration between app stores, payment platforms, and AI labs essential to disrupting these networks.