AI Is Getting Really Good at Messing With Cybercriminals
By Lily Hay Newman and Matt Burgess | Security | October 10, 2026
Anti-cybercrime initiatives are increasingly using AI to scam the scammers—tricking them into conversations with lifelike bots that they believe are real victims.
The Rise of AI-Powered Scam Baiting
As 2026 progresses, a growing number of cybersecurity organizations and law enforcement agencies are deploying generative AI to disrupt criminal operations from the inside. The strategy flips the traditional scam model on its head: instead of protecting victims after the fact, these systems actively engage fraudsters, wasting their time, extracting intelligence, and undermining their business models.
From Honeypots to Conversational Traps
Classic honeypots—decoy systems designed to lure attackers—have evolved. Modern AI-powered honeypots can hold natural, multi-turn conversations across text, voice, and even video. They mimic vulnerable targets: lonely retirees, eager job seekers, or distracted online shoppers. The bots respond with human-like hesitation, typos, and emotional cues, keeping criminals engaged for hours or days.
According to researchers presenting at security conferences this year, these systems are now capable of:
- Scalable engagement: A single AI can manage thousands of simultaneous conversations, far beyond human capacity.
- Real-time adaptation: Models adjust their persona based on the scammer’s tactics, language, and pressure points.
- Intelligence extraction: Every interaction feeds data on criminal infrastructure, payment methods, and scripts back to analysts.
Why Criminals Are Falling for It
Scammers rely on volume and efficiency. They cast wide nets, expecting most targets to ignore them. AI bots exploit that expectation by appearing to be perfect marks—just gullible enough to keep the conversation alive, but never actually sending money. The result is a drain on criminal resources: time, attention, and operational bandwidth.
“The tables have turned,” one European law enforcement official told WIRED. “We’re no longer just chasing them. We’re making them chase ghosts.”
A New Front in the Cybercrime Economy
The economic impact is real. Underground forums in 2026 are filled with complaints about “wasted leads” and “AI traps.” Some criminal groups have begun training their own models to detect bots, sparking an arms race in conversational AI. But for now, the defenders appear to have the upper hand.
What’s Next
As voice cloning and video synthesis improve, the next generation of anti-scam AI will likely be indistinguishable from real humans over any medium. Ethical questions remain—particularly around consent and deception—but for now, the consensus is clear: if AI can be used to defraud, it can also be used to defend.
This article is part of ongoing coverage of AI’s role in cybersecurity and digital fraud prevention.
via Wired AI