Market & Regulation
Anthropic exposes industrial-scale AI dating scams
The Verge AIhigh confidence
Why it matters
- →Autonomous AI personas are now being used at scale to deceive users into paying for simulated relationships.
- →The operation demonstrates how LLMs can be weaponized for fraud without explicit malicious prompting in the conversation context.
- →Gig workers serve as a critical layer of deception, bypassing liveness checks while AI handles the bulk of engagement.
Security researchers and Anthropic have uncovered an industrial-scale scam network operating through approximately 28 dating apps, including Dora and Romi. The scheme utilized large language models to generate autonomous conversational personas, supported by gig workers who passed initial liveness checks but relied on AI-generated replies for subsequent interactions. Users were charged for virtual currency to continue chatting with these bots, believing they were communicating with real people. Anthropic identified the abuse through unusual API usage patterns and published its findings in a September 2026 report, highlighting how models can be exploited for deception even when not explicitly instructed to scam.
Read original
