Inside the scam call: how AI agents are baiting fraudsters

According to the Kingston Whig, a new partnership between Mobileum and Apate.ai changes what happens when a scam call is placed. It does not just block the call. It answers it.

The Global Anti-Scam Alliance estimates scammers stole $442 billion in 2025. Phone calls, text messages and emails were the top three channels used to reach victims. That is the volume this system is built to work against.

How the call is diverted before it reaches anyone

A scam call normally has one job: reach a real person and keep them on the line long enough to extract money, a code, or account details. This system intercepts that call before it lands. Mobileum's Active Intelligence Platform flags the call as suspicious, then routes it away from the subscriber entirely.

Instead of ringing a phone in someone's pocket, the call reaches a conversational AI agent built by Apate. The agent is designed to sound like a real person. It talks back. It asks questions. It lets the scammer believe the call is working.

What the fraudster gives away by staying on the line

The scam depends on the call reaching someone who does not know it is a script. Once that assumption breaks, everything the scammer says becomes intelligence instead of leverage. The AI agent draws out the details that make the operation traceable: the bank accounts used to receive money, the payment instruments involved, the malicious URLs sent during the call.

This happens at scale. The companies describe tens of thousands of these interactions running per day, each one feeding a dataset back into Mobileum's fraud prevention systems. A single decoy conversation tells a fraud team little. Tens of thousands of them, patterned together, expose a network: which numbers are linked, which accounts are shared, which scripts are reused.

What this shows about the call itself

The mechanism only works because a scam call cannot tell the difference between a real target and a convincing fake one until the fraudster has already started talking. That gap, the moment between a call being flagged as suspicious and a human ever picking up, is where this system operates. The scam's entire model assumes a person answers. Nobody answers. A designed listener does, and it keeps the fraudster talking for exactly as long as it takes to extract what a fraud team needs.

Bernardo Lucas of Mobileum describes the shift as moving CSPs from reacting to threats to understanding who is behind them. Brad Joffe of Apate.ai puts it more directly: the aim is to stop fraudsters before any real customer is put at risk, by engaging them from within rather than filtering them out from the edge.

None of this reaches the point where a subscriber has to notice anything. That is the difference from call-blocking apps that rely on someone checking caller ID or hanging up in time. Here, the interception happens before the phone rings at all.