Technology

Fake Dating Apps Used AI Girlfriends to Scam People

Martin HollowayPublished 2month ago3 min readBased on 2 sources
Reading level
Fake Dating Apps Used AI Girlfriends to Scam People
Image by JESHOOTS-com from Pixabay

Anthropic traced a prepaid account sending more than 100,000 requests a day to its Claude AI system to a network of about 28 fake dating apps run by AI personas, fake profiles that chat on their own. The finding was described by Anthropic threat intelligence researcher Chris Cronbaugh, and shifts romance fraud from manual catfishing by individuals to automatic scams run at large scale. The Verge

Cronbaugh presented the operation in a talk titled "Swipe Right, Pay Up: Industrial-Scale AI Catfishing" at the Sleuthcon cybersecurity conference. Anthropic published its findings in the scams and fraud section of its "Detecting and countering misuse of AI: September 2026" report.

The network used AI personas to talk with users across the 28 apps. Most chats involved no human at all. Like a factory line, the software handled first contact, small talk, follow-up messages and escalation without a person stepping in.

Only one in four matches were real people. Those people were paid gig workers, not dating users. They were kept in the process to pass trust checks that software alone still fails.

The gig workers were hired to pass liveness checks on video, short live tests meant to prove a real person is present. Once past that check, they did not write their own messages. They answered by picking from three prewritten replies. That limit kept replies consistent, cut down language mistakes and meant less skilled operators were needed.

Security researcher Matthew "Zigula" Gore-Kormanik ran into the system while studying Dora, one of the fake apps. During his work he got a call notification from a persona named Jennifer.

Jennifer's bio described her as a 41-year-old Sagittarius with red hair, blue eyes and piercings. It listed interests in music, horror movies, nightlife and sports. The profile looked normal. Nothing in the background data attached to the profile gave away automation.

When Gore-Kormanik answered the call, his video screen showed a moving tapestry instead of Jennifer, with garbled background sound. No human was on the other end. The call acted like a broken video-call connection tied to a scripted profile.

After the call, Jennifer messaged Gore-Kormanik saying his voice was better than expected, even though his microphone was not connected. That comment was impossible in context. It showed the voice, video and chat parts were not keeping track of the same events.

McAfee had described a wider pattern in Valentine's Day 2026 research on romance scams across dating and social apps, which found AI bots and fake profiles driving fraud. McAfee

The broader context here is operational, not just technical. The usage logs held by the AI provider gave the first clue. High request volume, prepaid billing and very similar prompts are easier to spot at the provider level than fake profiles are to spot at the app level. For defenders who run AI systems, limits on request numbers, asking customers to state their planned use and grouping accounts that behave the same way matter as much as reviewing bad content.

Looking at what this means for proving identity, a video check alone no longer settles who is present. The workers passed the video check, then the model took over the conversation. That split, a human for the ID proof and a machine for the talking at scale, points to a need for ongoing checks tied to behavior during a session rather than a single check at sign-up. The long-term edge still sits with defenders who can link identity, device, network and language clues together, because large-scale fraud leaves large-scale traces.