Sydney start-up’s AI bots turn tables on phone scammers

He asks the man if he authorised a purchase of $999.98 for a MacBook to be delivered to an address in San Diego, California.

“Are you for real, mate? I’m in Sydney, not bloody California,” responds the man taking the call.

From there, the conversation continues along a familiar path. The Sydney man sounds confident but a bit naive – and he’s taking every hook from Dayton.

Dayton asks him to verify his credit card number. This takes time as the person on the phone fumbles to find his Visa card.

“Sorry, mate, it’s not the greatest of days for me,” he confides.

The conversation goes back and forth. At every step, Dayton thinks he is close to getting the credit card details, but never does.

That’s because only one of the two people in the conversation is real. In this instance, there was no “man” from Sydney, nor is there a credit card. “He” is an AI bot based in a data centre and holding a convincing two-way conversation with a likely fraudster.

Turning the tables

As the two talk, the bot is soaking up details and nuances of Dayton’s scam, while offering just enough to keep him going.

The scammer has been taken out of action, prevented from targeting a real victim.

The oldest trick in the book now has a high-tech twist: it’s the scammers being scammed.

This AI bot has relatives: male and female, young and old, Aussie, American, Indian and Chinese. All these “victims” are taking calls, holding conversations and keeping the scammers occupied while drawing out information.

This story is from The List: Australia’s Top 100 Innovators of 2025, published online and in The Australian on October 10.

The technology powering these bots was developed by Apate.ai, a Sydney start-up that was appropriately named after the Greek goddess of deception, and is emerging as a cutting-edge tool in the multibillion-dollar fight against global fraud. (In Australia, about $2bn is estimated to be lost to scams annually.)

It is already being used by Commonwealth Bank and local telco TPG, and is now being rolled out to global enterprises, including European telco giant T-Mobile.

Like all good technology, it started with a simple idea. Apate founder and CEO Dali Kaafar got a scam call while having lunch with his family.

But the caller walked into a trap, having called the executive director of Macquarie University’s Cyber Security Hub.

“I thought, ‘I’ll just have some fun with the kids’,” Professor Kaafar recalls.

“The whole thing went as a full comedy show for the kids, who wouldn’t stop laughing.

“I was pretending to be vulnerable and confused. I ended up spending 44 minutes with a scammer, just by playing along.

“As I hung up, I thought to myself, ‘If I could do that just for fun, surely technology could do it much better’.

“My wife then reminded me those were 44 minutes of my life I will never get back.”

Setting the trap

The next day at Macquarie University, Kaafar gathered a handful of his PhD students and prompted them with an idea: How can we automate this deception to provide protection to unsuspecting victims?

This is the key to cleverly using technology: performing a laborious task to a high level and applying it at scale. Ultimately, it offers a solution to a widespread problem.

For Kaafar, the concept is simple: every minute spent talking to AI is a minute the human scammer is not talking to a real victim.

“It’s about what scammers have as their most valuable asset: their time and the human resources they’re putting into this,” Kaafar says.

“You’re essentially breaking their whole business model.”

An equally important task is to mine the scammer for information, enabling real-time identification of scams as they evolve to prevent a scam propagating.

Kaafar researched how humans converse, including how people transition from one topic to another.

With that work as a foundation, the AI bot has a “roadmap” to conduct conversations.

Training the AI models involved intensive seeding of publicly accessible information about the concept of scamming.

There was painstaking training into how different types of scams work, from ATO scams to obtaining credit card details.

But the key aim of the bots was to hold scammers on a call for as long as possible. Kaafar’s team also made the bots aware of one another through benchmarking performance.

“We’d tell the bots [that] a good performance means a very long duration of conversation with each scammer, [and] they’d compete against each other,” he says.

“If they underperform compared to the average, reinforcement learning concept means they will look at themselves and learn from it.”

The technology allows suspect inbound calls to be automatically intercepted by the Apate bots, but is not just limited to voice; it also uses text-based platforms, such as WhatsApp and SMS. Apate’s early work with TPG has already diverted more than 280,000 scam calls, amounting to more than 100 days of scammers’ time wasted and a potential saving of millions of dollars in customer losses.

James Roberts, Commonwealth Bank’s general manager of group fraud, describes Apate as “flipping the script” against fraudsters.

“Every minute a scammer is engaging with a bot is a minute they’re not targeting an Australian,” he says. “The near real-time intelligence being gathered is a game-changer in how we help protect our customers and the broader community.”

Apate had early backing from Canberra’s Office of National Intelligence, and has since been backed by venture capital fund OIF Ventures. In August, it successfully raised $2.5m in a new funding round.

If Apate can train its AI bots against scammers, surely scammers can unleash their bots on us? It’s already happening, Kaafar says, albeit slowly, making the technology Apate is developing even more critical.

“A bot-versus-bot situation is a perfect situation where we humans will be essentially shielded against this whole cyber world,” he says. “We will be out of reach.

“We’re trying to fight the scammers outside the battlefield, and bringing it to a world where humans are not part of it. Which is what we want – to take it very far from the actual victims.”

This article first appeared in The Australian as Sydney start-up’s AI bots turn tables on phone scammers