Updated
Updated · The Verge · Sep 16
Anthropic Exposes 28 AI Dating Scam Apps That Hit 25,000 Users
Updated
Updated · The Verge · Sep 16

Anthropic Exposes 28 AI Dating Scam Apps That Hit 25,000 Users

3 articles · Updated · The Verge · Sep 16

Summary

  • Anthropic said a scam network used about 28 dating apps to lure users into paying for chats that were largely handled by AI personas, not real matches.
  • Over a two-week period in April, the operation deployed more than 4,700 fabricated personas, sent roughly 2.36 million messages, and engaged at least 25,000 people after one prepaid Claude account began making 100,000-plus API requests a day.
  • Only about 1 in 4 apparent matches was a human gig worker, Anthropic said; those workers mainly passed liveness checks while backend systems faked likes, visitors and video to keep users buying coins.
  • Most of the apps have since been removed from Apple and Google stores, but Kira was still live on Google Play as of Sept. 16, despite Anthropic saying it shared findings with both companies.
  • Researchers linked the network to a China-based actor through leaked Chinese-language internal materials, shared code and infrastructure, underscoring Anthropic's warning that disrupting such scams requires app stores, payment firms and AI labs to act together.

Insights

How did an AI scam network bypass Apple and Google's defenses to manipulate thousands before anyone noticed?
Why did a sophisticated AI dating scam still need paid human gig workers to successfully deceive victims?