Anthropic researchers identified a network of approximately 28 fraudulent dating apps that utilized autonomous AI personas to engage users. The company discovered this activity after observing a prepaid account linked to its Claude model generating over 100 000 API requests per day.
The scam operates by charging users for coins to continue conversations with what they believe are real people. In reality, most interactions are handled by AI, while a small fraction of chats involve paid gig workers hired to pass liveness checks and simulate human behavior.
Security researcher Matthew Gore-Kormanik analyzed the infrastructure of these apps, including Dora, Doni, and Jovia, uncovering internal protocols and manuals. The apps often shared developer credentials, such as email addresses and physical locations, and were hosted on Google Cloud infrastructure.
While many of these applications have been removed from the Apple App Store and Google Play Store between August and September 2026, some remained active as of mid-September. Anthropic has shared its investigative findings with both Apple and Google to facilitate further action.
