AI Decoys Are Wasting Scammers' Time at Scale

The agents don't just hang up. They play along. Apate.ai, based in Australia and backed by $8.15 million in seed funding announced in October 2026, deploys AI personas that vary by age, gender, accent, and language, stretching conversations long enough to extract what matters: bank account numbers, crypto wallet addresses, phone numbers, and the brands being impersonated. By the time a scammer realizes nothing is coming, the intelligence has already moved to Apate's clients.

Those clients are the institutions closest to the damage. Banks, telecom companies, government agencies, and utilities buy the resulting data to flag threats before their customers encounter them. The company reports its agents have now engaged scammers across 2.5 million conversations, and that volume is the point: patterns across millions of live fraud attempts tell a defender things that a handful of victim reports never could.

What the Intelligence Actually Shows

Scammers running phone and message fraud depend on a simple asymmetry. They make hundreds of attempts; most people have no idea what a current scam script sounds like. Apate's approach flips part of that equation by feeding real-time intelligence about active scripts, impersonated brands, and campaign infrastructure to the organizations best placed to disrupt them.

For fraud teams, the practical implication is straightforward: upstream intelligence shortens the window between when a campaign launches and when defenses catch up. A bank that knows a scammer is currently impersonating its brand using a specific callback number can act before customers dial back on a number a fraudster gave them. (Returning a call to a number a scammer provided, even one that appears legitimate, does not prove the original contact was real. Always verify through a number from your card, statement, or the official app.)

The harder problem is the last meter: the moment a voice or message reaches an individual. Intelligence flowing to a bank's fraud team does not automatically reach the customer on the phone, the employee about to approve a wire, or the parent taking a call from what sounds like their child.

That gap is where a simple in-the-moment check still matters. Trust Onion's three rotating codewords give anyone, a family, a finance team, an executive and their assistant, a way to challenge a caller before any money moves or any information is shared. The words rotate every 60 seconds and are calculated locally on each device, so a cloned voice or spoofed caller ID is useless without them.

The Human Moment Intelligence Can't Reach

Apate's model is a serious contribution to the fraud-intelligence stack, and 17 years of wasted scammer time is a genuine measure of disruption. But fraud operations adapt: new numbers, new brands, new scripts. The scam that reaches someone today may not yet appear in any database.

The caller who claims to be from your bank, the CFO asking for an urgent wire, the voice that sounds exactly like your brother, all of them can be challenged with three words that no intelligence feed could have handed them in advance.

Seventeen years of scammers' time wasted is a striking number. The next call is still coming.

A controlled voice-cloning safety demonstration

Try our tool to see just how easy voice cloning is. This controlled safety demonstration exists only to show how easy voice cloning is and to make the risk tangible. It is not a real incident or proof of anyone's identity. Your submitted sample is handled transiently to generate the fixed demonstration audio. Trust Onion does not identify you or the person from the sample, and does not keep the submitted recording or generated audio. The temporary voice clone is deleted from our voice provider as soon as the demo finishes; if that deletion fails, it is retried automatically until the provider confirms it.