Anthropic Exposes AI Catfishing Network Across About 28 Dating Apps

The scam is not a conventional romance fraud in which a fake partner asks for emergency money or cryptocurrency. The apps themselves are the paid product: users buy coins with real money to keep talking, mostly to machines, while believing they are talking to other people. Many of the apps were still live on both major US app stores at the time of the reporting, even after Anthropic’s public talk.

Security researcher Matthew “Zigula” Gore-Kormanik saw how the system worked while analyzing Dora, one of the fraudulent apps. He received a pop-up saying Jennifer was calling. Her profile described her as a 41-year-old Sagittarius with red hair, blue eyes and piercings who liked music, horror movies, nightlife and sports. Gore-Kormanik answered, but Jennifer did not appear in the video feed. He saw a tapestry moving, probably because of a fan, and heard distortion in the background. After the call, “Jennifer” messaged him to say she had had fun and that “your voice is way better than expected.” His microphone had not been connected.

Gore-Kormanik examined Dora and similar apps after a Verge reporter shared a spreadsheet of potential dating scam apps with him. On June 5, Anthropic let Chris Cronbaugh, a threat intelligence researcher at the normally tight-lipped company, give a talk at the public cybersecurity conference Sleuthcon. The talk was called “Swipe Right, Pay Up: Industrial-Scale AI Catfishing.” Cronbaugh described a network of around 28 dating apps in which conversations were largely run by autonomous AI personas. The majority of chats did not involve a human agent at all, he said.

Anthropic later published the findings in the “scams and fraud” section of its “Detecting and countering misuse of AI: September 2026” report, after The Verge had spent extensive time looking into the network and trying to corroborate the findings. Anthropic said it uncovered the fraudulent apps after noticing unusual activity on Claude: a prepaid account sending over 100,000 API requests per day.

The apps presented themselves as ordinary dating services for real people. Dora was described as “a dating app thoughtfully designed for wide range of ages people … a respectful easy-to-use space to meet people who share your values,” while another version called it a “warm, simple dating app for adults seeking real connection.” Romi said it “helps you discover, connect, and chat with real people.” Doni’s tagline was “start real companionship,” and its description said it “helps you connect with nearby singles.” They did not look like companion apps such as Replika, where the personas are clearly AI.

According to Anthropic, the users were mostly men in their mid-to-late 30s. They matched with profiles they believed belonged to real women, but only one in four of those people was real, and that person was not a user looking for dates but a paid gig worker. The gig workers were hired to pass liveness checks on video or to follow social media accounts. They did not write their own comments; instead, they responded to messages by selecting from three pregenerated replies. The workers could prove they were human, while the AI personas could only offer plausible explanations for why a video chat or call was not possible.

Anthropic’s report said backend components fabricated likes, visitors and pre-recorded “video” when no real person was available, and tracked which users had begun to suspect they were talking to a bot. Because users occasionally interacted with real gig workers, some began to believe that the entirely AI-generated replies they received on the app were also from real people.

Multiple AI providers were involved. Claude was misused to run the autonomous conversational personas, according to the report. A small non-Anthropic model generated the short reply suggestions that gig workers tapped, along with face-attractiveness scoring and photo and voice moderation. An image-editing model generated avatar imagery.

The scheme sits within a broader ecosystem of online scams, including automated thirst trap replies, fake social media accounts and fake dating app profiles. In many romance scams, people spend months building trust before claiming an emergency and asking for financial help, or they pretend to be investors and ask for money in an account they control that turns out to be fake. This network avoided those tactics. There was no money “borrowed” by fake romantic partners, and no cryptocurrency was involved. Instead, the apps charged for coins needed to continue interactions, and the coins cost real money.