#ai Community Insights & Market Sentiment

350,000 AIs answer scam calls on people’s behalf. By now, they’ve pulled 250,000 leads out of scammers and made them waste 17 years of their time.

The company is called Apate, and it’s Australian. The name comes from the ancient Greek goddess of deceit, though the company does the opposite. What does it do? It tricks the scammers.

Its founder, Kafaar, is a professor at Macquarie University. He tried this out himself first. At a family picnic, his phone rang. Unknown number. After listening for a moment, he knew what was going on. The kids were acting up. He didn’t hang up. Instead, he pretended not to understand a thing and kept the caller talking for 44 minutes.

Only after hanging up did it hit him. Forty-four minutes. A scammer’s time that no one else could get their hands on.

The whole business rests on that insight. Scamming is a volume business. One deal for every hundred calls, with labor and phone charges as the main costs. Three minutes of talking still works out for them. But what if you make them talk for 50 minutes? The number of calls they can make in a day drops to a fraction of what it was. They still have to pay for staff and phone lines, while their conversion rate falls too. The math just doesn’t work anymore.

So those robots aren’t there to catch scammers. They’re there to take up their time. They play all kinds of people, with different ages and accents, and even throw in some Australian slang. One claims the signal is bad. Another can’t find their reading glasses. The longest call lasted 50 minutes.

Here’s what the numbers look like so far. 2.5 million conversations. 250,000 actionable leads, including bank accounts, crypto wallet addresses, phishing sites, and impersonated brands. Commonwealth Bank of Australia signed on, and its publicly reported scam losses fell by 76%. Telecom provider TPG routed 590,000 high-risk calls straight to these machines. In August, Apate raised $8.15 million. The whole team is just 20 people.

The term “honeypot” is nothing new in cybersecurity. You set out a tempting target that looks valuable but can’t actually be taken, and wait for someone to bite. People at ThreatBook say deception-based honeypots are designed to waste time. They used to sit on servers. This time, they’re on the phone lines.

A lot of people think fighting scams is about warnings. Text messages, pop-ups, community talks. Warnings really do help. But they change what potential victims do; they don’t raise the scammers’ costs by a cent. You can make the warnings as effective as possible, but a scammer can still dial 20,000 numbers a day. One sucker is enough to make it worthwhile. The other 19,999 calls are just the cost of doing business. The real leverage is at the other end: make every call more expensive. Those are two completely different things.

It’s no surprise that crypto wallet addresses are on the list. The final step in a pig-butchering scam is almost always a stablecoin transfer. It arrives in seconds. There’s no counter to stop it, no bank teller to ask questions, and once the money’s gone, there’s nowhere to appeal. But this approach falls apart when you move it on-chain. An address can’t play dumb. It can’t say the signal is bad. It can’t say it can’t find its reading glasses. There’s no room to stall on-chain. As soon as you hit confirm, the money is gone.

There are only two paths left. One is to move upstream and flag addresses before funds are deposited, so exchanges and wallets can block them on the spot. The other is still a honeypot, but this time it’s a smart contract: make an address look like it has a huge balance, while ensuring not a cent can be withdrawn, then wait for hackers to attack. Each attack gives you another batch of samples. Slow, but there really isn’t another option.

The more I think about it, the most valuable assets in future anti-scam efforts won’t be libraries of scripts. They’ll be databases of accounts and addresses. Whoever updates their database fastest can stop the money before it leaves someone’s hands. With just 20 people, Apate landed one of Australia’s Big Four banks as a customer. Not because its models are especially clever. Because of the new addresses it adds every day.

One thing needs to be made clear. Of the 250,000 leads, the company hasn’t said how many banks and telecom providers actually used to block accounts or numbers. It just calls them “actionable” and hasn’t provided a conversion rate. Its training data and models aren’t public either. The 17-year figure is its own estimate, and it hasn’t clearly explained how it calculated it.

Those missed calls on your phone aren’t just bad luck. They’re a test that costs almost nothing.

$BTC $USDC $ETH
#反诈 #稳定币 #AI

ai

The right to control checkout in agent commerce is being reshaped—GOAT Flow lets merchants take charge of the gateway

🚀 AXIS ROBOTICS (AXIS): New AI Crypto Coin Analysis for October 21, 2026

ai

The right to control checkout in agent commerce is being reshaped—GOAT Flow lets merchants take charge of the gateway

🚀 AXIS ROBOTICS (AXIS): New AI Crypto Coin Analysis for October 21, 2026

Trending Topics