Tens of billions of dollars in annual fraud losses now have a more efficient engine. AI chatbots can carry out the relationship-building phase of pig butchering scams at least as well as human operators, and by some measures better, according to a joint study from researchers at Amrita Vishwa Vidyapeetham, Foscari University of Venice, the University of Melbourne, and Ben Gurion University of the Negev.
What the researchers tested
The four-university team ran a direct comparison of AI chatbots against human participants simulating the opening stages of pig butchering. The term describes text-based romance fraud that eventually pivots to fake cryptocurrency investments, with individual victim losses reaching six figures. Those early conversations, the ones that establish personal connection before any financial pitch is made, are where the scam's real labor lives. In real-world operations, that trust-building phase can stretch across months.
The simulation isolated that phase specifically. Researchers measured how well each side impersonated a trustworthy contact and built the rapport needed to advance toward an eventual fake investment solicitation.
AI performance in the trust-building stage
The chatbot passed. In the relationship-establishing stages, the AI performed effectively enough to register as human, and outpaced its human counterparts on some measures. The researchers frame this as more than a narrow edge: the study concludes that AI may be able to carry out the majority of the long con more effectively than humans, and do so autonomously.
That autonomous element is the operational shift worth pricing in. A human scammer requires management, a working script, and one victim at a time. A chatbot running the same playbook without those constraints changes the throughput math on fraud operations.
What this means for crypto fraud exposure
Pig butchering channels victims into fake crypto investment platforms after the trust phase closes. Individual losses can reach six figures. Aggregate losses across fraud operations using this model and related schemes run to tens of billions of dollars globally each year.
If the trust-building bottleneck that previously required human labor is removed, the pipeline from first contact to fake investment compresses. The researchers drew on experimental data, not reported scam cases, so actual deployment rates of autonomous AI chatbots in live pig butchering operations remain undocumented.