Beware, scammers. O2 has created a new weapon in the fight against fraud: an AI granny that will keep you talking until you get bored and give up.

O2, the mobile operator arm of British telecommunications giant Virgin Media (VMO2), says it has created human-like AI to answer calls from fraudsters in real time, keeping them busy on the phone and wasting their time pretending to be a potential vulnerable target.

“Daisy” is claimed to be indistinguishable from a real person, fooling scammers into thinking they’ve found the perfect prey thanks to her ability to engage in rambling “human-like” chatter, the company claims.

In the weeks leading up to International Fraud Awareness Week (November 17-23), AI has already frustrated scam callers with meandering stories about her family and spoken at length about her passion for knitting, according to O2.

At this point, many Reg readers probably feel they know someone the telco would have used as training data.

But phone scams are an increasingly common threat. Criminals, often working out of call centres, cold call lists of numbers to try to scam people out of their money. Common tricks include pretending to be their bank or a courier who needs payment to deliver a package to get them to divulge their bank account details.

Daisy is said to combine various AI models that work together to listen to fraudulent calls and respond immediately, as if engaged in a conversation. Appropriate responses are generated by a custom large language model (LLM) with a character “personality” layer, and then fed through a custom AI text-to-speech model to generate the spoken response.

O2 claims it’s so realistic it’s successfully kept fraudsters distracted for up to 40 minutes at a time. Some scammers have even been tricked by Daisy into providing false personal information, including made-up bank details.

Reg I wondered how the telco managed to make this work in practice. Does Daisy intervene if a call to a user on its network comes from a number suspected to be a scam operator?

Unfortunately, it’s not that sophisticated. AI has its own dedicated number, which the anti-fraud team managed to infiltrate into contact lists used by scammers to target Britons, an O2 spokesperson told us.

“She has her own number. We worked with top scammers as we developed Daisy, including the UK’s biggest scammer, Jim Browning, and they gave us guidance and advice to help us attract scam calls,” the spokesperson said.

Jim Browning is the online alias of a software engineer and YouTuber from Northern Ireland who appeared in a BBC Panorama documentary about the fight against fraud.

“Working with Jim and using a range of tactics, including something known as number seeding, we were able to get Daisy’s phone number added to a list of online ‘mug lists’ used by scammers. Then we just had to wait for the calls to come. Inside, with Daisy primed and ready to answer whenever the phone rings 24/7,” the spokesperson explained.

While it doesn’t directly protect people from scam calls, O2 believes that by wasting the scammers’ time, it has diverted them from targeting real victims, while providing insight into the tactics fraudsters are trying to use against their targets. And it’s good to hear about a useful application of AI for a change.

“We’re committed to playing our part in stopping fraudsters, investing in everything from firewall technology to block scam texts to AI-based spam call detection to keep customers safe,” said VMO2 fraud director Murray Mackenzie , in a provided remark. .

Now if only O2 could come up with ingenious solutions to solve other problems such as customer service waiting times. ®