A new study suggests AI chatbots can produce “exploitable trust” better than human scammers, at least in the part of pig butchering fraud that typically takes the longest: the months-long relationship conversation.

The researchers, coming from Amrita Vishwa Vidyapeetham (India), Foscari University of Venice, the University of Melbourne, and Ben Gurion University of the Negev, ran an experiment where AI chatbots were pitted directly against humans in a simulated scamming process. Their focus was not on the final ask for money, but on the trust-building steps that lead to soliciting a fake investment, often framed in text-based romance that later shifts into fake crypto pitches.

So why does this matter to executives, boards, and anyone responsible for fraud risk? Because pig butchering is not a “spray and pray” scam. It is social engineering at scale, a methodical long con. In the real world described in the source, victims can lose as much as six-figure sums, and scam operations steal “tens of billions of dollars a year worldwide.” That sets the economic stakes, but the simulation sets the operational ones: if the bottleneck is “building trust,” and AI is better at that than humans, then the cost to run scams drops while the success rate can rise.

The study’s experiment is built around a realistic sequence: the relationship-establishing stage, the stage in which scammers impersonate someone, build familiarity, and gradually guide the victim toward financial participation. According to the source, the relationship-establishing portion often stretches to months. In the researchers’ simulation, an AI chatbot performed remarkably effectively in impersonating a human and, by some measures, outperformed the real human “scammers” in that trust-building phase.

There is an uncomfortable second-order implication here for anyone trying to contain fraud through detection alone. Traditional defenses often look for signals that appear late in the funnel, like explicit investment solicitations or obvious links to fake platforms. But the source points out that the most time-consuming part can be trust cultivation. If AI systems can shorten, accelerate, or improve that cultivation, then the “early warning” window is the conversation itself, not the final transaction. That changes what product teams need to monitor, what investigators need to triage, and what risk models should prioritize.

It also complicates platform accountability and regulatory framing. Fraud enforcement and consumer-protection regulators are increasingly dealing with generative AI as a general-purpose capability. The source is careful not to claim AI fully replaces human scammers autonomously building an entire web of deception end to end. Instead, the study’s signal is narrower but more dangerous: AI seems to do very well at the long trust-building conversations that set up the ask. In other words, even partial automation can be enough to tilt outcomes. Regulators and compliance teams typically do not need proof that a whole crime is automated to act. They can act on the measurable increase in effectiveness, especially when victims are being driven toward fake crypto investments that can total six-figure losses.

From a governance standpoint, boards should treat this as a “capability arms race” issue, not just a customer-safety issue. Scam operators and legitimate organizations both benefit from language tools. When scammers get better at producing believable, relationship-driven messaging, it becomes harder for humans to detect deceit and harder for systems to flag content that is still in the “friendly talk” phase. That puts more pressure on data strategies, reporting workflows, and incident response playbooks. It also increases the likelihood that fraud prevention must move upstream in the user journey.

If you are a CEO, CFO, or security leader at a company that hosts messaging, social interactions, or online financial activity, the strategic stake is straightforward: a shift in who can generate trust faster and more convincingly shifts fraud ROI. The study indicates AI chatbots can outperform humans in a key step of pig butchering, the long con that often leads to fake crypto investments. In a world where the fraud industry is already stealing tens of billions annually, better trust generation is not a technical curiosity. It is a new lever for attackers, and it forces defenders to reconsider how quickly they can detect deception before money moves.