Scam-Baiting Bots: Anti-Fraud Teams Turn AI on Cybercriminals

Anti-cybercrime operations are increasingly pointing AI back at the people who run fraud campaigns, using lifelike bots that impersonate real victims and keep scammers engaged in conversation. The approach flips a familiar dynamic: instead of a human baiting a fraudster, software carries the exchange, answering messages and staying in character for as long as the criminal keeps typing.

The mechanics are straightforward. A bot adopts a plausible identity, responds at human speed, and feeds the scammer just enough interest to continue without ever handing over money, credentials, or a working payment route. Because the persona never tires, one operator can run many conversations at once, turning what used to be a volunteer pastime into something closer to an automated operation.

That shift matters for the economics of fraud. Scam outfits work on volume, and every hour a criminal spends on a dead-end conversation is an hour not spent on a real target. Baiting has long been a volunteer pursuit, but it depended on people with spare time and thick skin. Software removes both constraints, which means the tactic can scale in a way individual hobbyists never could.

The same capability cuts the other way. The text-generation tools that make a convincing fake victim are the same ones that make a convincing fake bank, employer, or romantic partner. Anti-fraud researchers have spent years documenting how language models lower the cost of writing fluent, personalized lures in multiple languages, and the baiting wave is essentially that same toolkit aimed in the opposite direction.

There is also a measurement problem. A bot that talks to a scammer can log what the criminal asks for, which payment methods they push, and which scripts they reuse, producing data that is hard to gather from victim reports alone. Whether that intelligence reaches the banks, platforms, and law enforcement agencies in a usable form is a separate question, and one that will decide how much the tactic is worth beyond the entertainment of wasting a fraudster's afternoon.

Expect more of this as the tools get cheaper and easier to run, along with pressure to show it actually reduces losses rather than just occupying criminals. The open question is verification: how anyone proves a bot's conversation prevented a real theft, and whether scammers learn to spot the tells fast enough to make the whole exercise moot.