AI chatbot outperforms human scammers in building trust for romance fraud

Scholars at Amrita Vishwa Vidyapeetham, Foscari University of Venice, University of Melbourne and Ben Gurion University of the Negev investigated whether generative AI can replace human operators in "pig-butchering" romance scams. In a controlled week-long trial, participants conversed with both a seasoned human scammer and a Claude-based AI; the AI achieved a 46 % success rate in getting victims to download a test app, compared with 18 % for the human, and received higher trust ratings.

Victims sent 80 % of their messages to the AI and only one participant suspected it during the study. Researchers argue scammers could use large language models for the lengthy trust-building phase, handing over the final investment request to a human to bypass AI safeguards. Interviews with 145 former scam workers highlighted the reliance on forced-labor victims in Southeast Asian compounds, but AI may eventually reduce this labor pool. Anti-fraud advocates warn AI-driven scams could become more effective and harder to dismantle, even as they potentially lessen human trafficking.