Commonwealth Bank deploys AI bots with Aussie accents to trap scammers using ‘honeypot’ strategy

Telephone scammers have a new mark to contend with – an artificial intelligence bot deployed by Commonwealth Bank to tie them up in tedious, go-nowhere calls.

Phone scams accounted for the highest amount of financial losses in 2024, according to the National Anti-Scam Centre, and CommBank hopes this new tool will help protect Australian citizens.

“This is about flipping the script. Scammers are increasingly using AI to target Australians – we’re turning the tables by using AI to fight back,” said James Roberts, CommBank’s General Manager of Group Fraud.

“Every minute a scammer is engaging with a bot is a minute they’re not targeting an Australian.”

The technology used by CBA has been spun out of Macquarie University’s AI research lab and was trialled in a pilot program last year.

The spin out company, Apate.AI, is headed by Professor Dali Kaafar, who is also Executive Director of Macquarie University’s Cyber Security Hub.

Apate.AI uses thousands of conversational AI bots to disrupt scammers targeting Australians via text-based conversations and voice calls, tying them up in fruitless exchanges and sharing real-time information on scam types with the bank.

CommBank said it works with telecommunications companies, tech firms and government agencies to share intelligence and coordinate responses.

The system uses a “honeypot” strategy, operating a network of dedicated phone numbers designed to be discovered and targeted by scammers. Once the scammer calls or messages a honeypot number, they end up engaging with an AI bot instead.

“We’ve designed our bots to be difficult to detect by scammers, making them incredibly effective at gathering intelligence and disrupting scam operations. The bots are uniquely crafted with diverse identities – varying in gender, age, tone and cultural nuance – and fine-tuned with Australian slang and humour to improve realism,” Professor Kaafar said.

CommBank said it had diverted hundreds of thousands of scam calls since it began a trial program with Apate in 2024, and that the technology would become increasingly important as scammers also start using AI to power their nefarious acts.

“We know no single organisation or sector can solve this alone,” Mr Roberts said. “That’s why it is so important to collaborate across sectors to disrupt scam operations at scale.”

The initiative follows a similar effort in the UK that went viral late last year, when mobile phone operator Virgin Media O2 unveiled an AI chatbot named Daisy – a sweet-sounding virtual 78-year-old grandmother – to waste scammers’ time. Daisy was trained on real scam calls and designed to sound convincingly human, often keeping fraudsters on the line for upwards of 40 minutes with rambling stories about knitting and her cat, Fluffy.

Daisy spearheaded an advertising campaign to encourage Brits to report scam calls to a national hotline, with the promotional video views more than 1 million times on YouTube.

The Commonwealth Bank has been a fast mover in the adoption of artificial intelligence for its own business improvements. Earlier this year, the bank said AI was able to answer 90 per cent of customer inquiries and it was being deployed in the business bank to help reduce loan approval times to as little as ten minutes.

The bank has already deployed detect fraud, processing 20,000–30,000 fraud alerts daily, a number that could scale effortlessly to 50,000.

“The technology we have seen over the last six to nine months has the potential to be quite transformational,” Commonwealth Bank CEO Matt Comyn said at the time.