Anthropic's Cronbaugh: 28 dating apps ran AI catfishing at scale

The case came to light because Anthropic, a normally tight-lipped company, let its threat intelligence researcher Chris Cronbaugh give a public talk at Sleuthcon. Anthropic had noticed unusual activity on Claude: a single five-day-old prepaid account with no history suddenly started sending out 100,000+ API requests a day. That trail led to the dating app network, and the company published its findings in the "scams and fraud" section of its "Detecting and countering misuse of AI: September 2026" report — but only after The Verge had spent extensive time trying to corroborate them.

The operation was not a typical romance scam: there was no money "borrowed" by fake romantic partners, and no cryptocurrency involved. The apps themselves were the scam, charging for coins that users spent talking mostly to machines. Gig workers were hired to pass liveness checks or follow social media accounts, while AI personas kept conversations going 24/7. Claude ran the conversation layer, but the report says a small non-Anthropic model generated short reply suggestions for the gig workers, and an image-editing model generated avatar imagery.

The reporting also rests heavily on the work of security researcher Matthew "Zigula" Gore-Kormanik, who pulled the APKs, ran static and dynamic analysis, and found a leaked internal protocol manual shipped inside Doni. That manual documented monitoring of gig workers, call recording, transcripts, performance ranking, and the technique of pretending people had called users to lure them into conversation. Cronbaugh said accounts tied to the network had all been created in the previous month, underscoring how cheap it is to rebuild. The test for app stores and payment processors is whether they act on signals they already had, such as customer reviews, before the next network is built.