At first glance, Daisy seems like your typical grandmother – fond of knitting, chatting about her family, and slightly out of touch with technology. However, Daisy is anything but ordinary. Developed by British O2, she’s an AI-powered chatbot designed to outsmart phone scammers by engaging them in lengthy, time-wasting conversations.

Launched earlier this month, Daisy represents a creative use of AI to tackle fraud, which cost consumers worldwide over $1 trillion last year, according to the Global Anti-Scam Alliance. By keeping scammers occupied with human-like small talk, Daisy diverts their attention from real victims. O2 reports that she has held fraudsters on calls for up to 40 minutes, delivering a significant blow to their operations and providing protection to users.

Daisy’s voice and persona were inspired by a staff member’s grandmother, leaning into scammers’ biases by mimicking traits that might make her seem like an easy target. Her charm lies in her ability to share lengthy stories, provide fake personal details, and remain endlessly patient during phone calls.

“With scammers operating fulltime call centres specifically to target Brits, we’re urging everyone to remain vigilant,” said Murray Mackenzie, Virgin Media O2’s director of fraud. Last year, the company blocked over £250 million in suspected fraudulent transactions.

Using a large language model (LLM), Daisy interacts with scam callers in real time. Though she doesn’t intercept incoming calls, O2 circulates her phone numbers online to lure fraudsters into her conversations. While many Brits hesitate to engage with scammers themselves, Daisy has “all the time in the world,” keeping fraudsters tied up and frustrated.

As AI continues to shape both positive and negative aspects of online security, tools like Daisy showcase how technology can be leveraged creatively to protect vulnerable populations from digital threats.