Anthropic and security researchers have uncovered a widespread scam network involving approximately 28 dating apps that use autonomous AI personas to engage with users. According to a presentation by Anthropic's threat intelligence researcher Chris Cronbaugh at the Sleuthcon cybersecurity conference, the network utilized AI to conduct large-scale "catfishing" operations.
The scam operates by presenting users—primarily men in their mid-to-late 30s—with AI-generated personas. While some chats involved human gig workers who used pregenerated replies to pass liveness checks, the majority of interactions were driven by autonomous AI. Anthropic reported that the company detected the misuse after noticing a single prepaid account sending over 100,000 API requests per day via Claude.
The monetization strategy relies on users purchasing virtual "coins" to continue interacting with the personas, believing they are speaking with real humans. The network's infrastructure includes multiple AI providers; while Claude was used for conversational layers, other models were employed for image generation and reply suggestions.
Security researcher Matthew “Zigula” Gore-Kormanik identified that several apps, such as Dora, Romi, and Luma, were interconnected through shared code and backend architecture. Based on internal Chinese-language materials and infrastructure, Anthropic attributes the operation to a China-based actor. Although several apps have since been removed from major app stores, researchers noted that the highly organized nature of the operation allows scammers to quickly launch new applications to circumvent detection.
Sources
- The sexy AI-powered dating app scams are here （The Verge AI, 2026-09-16）