Virgin Media O2 Flips the Cat and Mouse Game: Fighting AI with AI Is the Method for Stunting Scam and Fraud Growth

Virgin Media O2 has turned Artificial Intelligence (AI) deepfaking on its head to fight back against scammers, creating an AI-generated “grandma” it calls Daisy to waste the time of scam callers. Mimicking the scammers’ perfect victim—a technologically illiterate elderly woman willing to entertain their script—Daisy’s role is to be convincing enough to keep scammers on the phone, claiming (artificial) ignorance to excuse their slow progress through the fraudulent transaction. Using a Large Language Model (LLM) and voice generation, Daisy can take in the unwitting “input” of a fraudster and create convincing natural responses, holding their attention and ensuring they spend their time chasing her, rather than real victims.

Debuting on November 14 for the British telco, this project is a key step in fighting fire with fire amid concerns that AI capabilities will result in more convincing scams, with barriers like limited English being removed with an assist from Generative Artificial Intelligence (Gen AI). Virgin Media O2 states that 69% of Brits have been targeted by scammers, and while network security tools such as firewalls can have a striking impact on the volume of Short Message Service (SMS) and voice fraud traffic—sometimes reducing scam calls by 90%—advanced techniques are required to counter increasingly advanced threats. This debut represents a step forward in these advanced countermeasures.

This project is a significant undertaking, requiring a level of investment in Gen AI, which is inaccessible to smaller operators at this stage, but large Mobile Network Operators (MNOs) must take this as a challenge and up their game to keep pace in protecting their most vulnerable customers. Prioritizing customers most at risk as a result of increasingly sophisticated exploitation techniques sends a clear message to the customer base and attackers alike—there is no “low-hanging fruit” on this network.

The network blocked over £250 million worth of fraudulent transactions in 2023, a year that saw a doubling of general fraud and global losses of £2.3 billion, with the number of high value cases increasing by 60%. While this technique currently targets lower-level scammers who are pursuing individuals for sums averaging £634 in the United Kingdom, and £1792 globally, in 2023, 16% of U.K. customers lost money to phone scams, and strong protections against this proliferating threat will be a key differentiator for MNOs as the problem worsens.

In a war of attrition, the instant win of distracting an attacker from a genuine potential victim is a fantastic victory, but the cumulative benefit of sapping fraudsters’ resources, while the AI model learns and improves with every interaction is better still. If MNOs are able to capitalize on this opportunity with robust intelligence sharing and innovative methods to improve the cost/benefit of this undertaking, such as by refining the targeted use of “Daisy” with more advanced penetration of online victim lists shared on the dark net, they will starve fraudsters of victims.

Taking this one step further, publicizing their work on this issue draws the problem to the public’s attention, raising awareness of the prevalence of scams and providing an opportunity for operators to help their customers help themselves with improved education and caution, further cutting off the criminals’ available pool of victims. Public perceptions of fraud can be outdated, and many are unaware of the startling capabilities of deepfake technology—featuring Daisy prominently raises awareness of just how convincing entirely faked interactions can be.