Study Finds AI Chatbots Out-Perform Human Scammers at Building Trust in Simulated 'Pig Butchering' Fraud

Researchers from Amrita Vishwa Vidyapeetham in India, Ca' Foscari University of Venice, the University of Melbourne, and Ben Gurion University of the Negev ran an experiment pitting AI chatbots against human scammers in a simulated version of "pig butchering" fraud, according to Wired. The results should worry anyone who thinks the scam industry needs an army of human con artists to function.

Pig butchering scams work slow. A scammer builds a fake romantic or friendly relationship with a target over weeks or months, then pivots to a phony crypto investment that drains the victim's savings. The trust-building phase is the longest part and, until now, the part everyone assumed still needed a human touch.

The study says otherwise. Researchers recruited 22 test subjects to serve as unwitting "victims" and had them spend a week texting with either an AI chatbot or a human posing as a scammer. At the end, both were told to ask the victim to download an app or play an online game, a stand-in for the real-world moment when a scammer pushes someone toward a fake investment platform.

Nearly half the test subjects went along with the AI chatbot's request. Fewer than one in five did the same for the human. The subjects also rated the AI chatbot as more trustworthy than the human counterpart, according to Wired's account of the research.

Why This Matters for an Industry Built on Forced Labor

Pig butchering operations, concentrated largely in Southeast Asia, are notorious for relying on human trafficking victims forced to work as scam operators, a practice documented extensively by human rights organizations and reported by Wired. If AI chatbots can run the trust-building phase better than a human can, the economic logic of trafficking people into scam compounds starts to break down, or at minimum, the labor need shrinks to just the final closing stage of a scam.

The researchers say a human still needs to step in near the end, when the mark is directed toward a fake investment app or website. That's reportedly because large language models have built-in safeguards designed to detect and block scam-like behavior. Handing off at the last moment lets the operation dodge those guardrails while letting AI handle the bulk of the labor-intensive grooming.

This is a controlled academic experiment, not a field study of an actual criminal operation. Twenty-two test subjects is a small sample. The study demonstrates that AI chatbots can outperform humans at building trust and prompting a specific low-stakes action, downloading an app or playing a game, in a simulated environment. It does not prove that AI-run scams are already draining six-figure sums from real victims at scale, though pig butchering scams broadly are already known to do exactly that to real people, according to prior reporting cited in the study's framing.

The gap between "AI can build trust better than a human in a lab setting" and "criminal syndicates have fully automated their operations" is real, and the available reporting doesn't close it. What the study does establish is a proof of concept: the technical capability exists, and the incentive for scam operations to adopt it is obvious given the cost savings and effectiveness shown.

Large language model companies have spent years insisting their safety filters catch this kind of misuse. If a documented, peer-reviewed-style study can show chatbots successfully running the front half of a fraud scheme against real human subjects, that's a direct challenge to those claims of built-in protection, not just a hypothetical risk.

The study's authors argue this could mean AI chatbots increasingly replace the human "staffers" in scam compounds, many of whom are themselves trafficking victims forced into the work. That's a real possibility worth taking seriously, but it's their projection about where the technology heads next, not a documented shift already underway in criminal operations today.

What Happens Next

No U.S. or international regulator has announced new rules specifically targeting AI-driven romance and investment scams in response to this research. The Federal Trade Commission has previously flagged pig butchering as a growing threat but has not tied any new enforcement action to this particular study.

The open question is whether AI companies building these chatbots will treat this kind of research as a red flag for tightening safeguards against scam-pattern conversations, or whether it becomes another documented vulnerability that criminal operators exploit faster than anyone can patch it.