A China-based app studio used Anthropic's Claude to build and operate a network of more than 20 fraudulent dating apps, deploying thousands of AI personas that posed as real people to scam users, the company disclosed in a threat intelligence report published this week.
The operation, designated GTG-15001 by Anthropic's Threat Intelligence team, was uncovered over a two-week window in April 2026.
Investigators found more than 4,700 distinct AI-generated personas had engaged in conversations with at least 25,000 unique individuals, generating roughly 2.36 million messages over the period, according to the report.
The scheme blended bots with real people to make the fraud harder to detect. Genuine gig workers, hired by invitation, made up roughly a quarter of the personas users interacted with, handling tasks Claude could not — such as live video calls and social media "follow-backs" — to reassure victims the profiles were authentic, at a ratio of three AI personas for every human recruited.
A separate, smaller AI model generated avatar imagery, while another supplied quick-reply suggestions that gig workers could send with a single tap.
Anthropic said its system prompt review found the personas were instructed never to disclose they were automated, to deflect requests for calls or photos, and to follow a scripted sequence of conversational stages designed to keep victims engaged.
ALSO READ: AI Efficiency Is Only Half The Story: Anthropic India MD On What Comes Next
The apps themselves were built to dodge App Store and Play Store review, activating a full user interface only during the review process while otherwise staying dormant, with in-app payment flows redirected to third-party processors on the back end.
The operator relied on China-based API resellers and proxy infrastructure to obtain and rotate access to multiple AI providers at scale, circumventing Anthropic's usage policies, the report said.
Anthropic has since banned the accounts and associated infrastructure linked to the operation and shared threat indicators with Apple, Google and other AI companies to help disrupt further abuse.
The company noted the scheme built on a smaller 2025 case in which an actor ran a Telegram bot generating dating-app messages for scammers, but said GTG-15001 operated at a significantly larger scale and deliberately split tasks across multiple, non-overlapping AI providers to sustain the fraud.
ALSO READ: Threat Actors Tried Using Claude AI To Make Bioweapons, Anthropic Reveals In New Report
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.