Playing dumb: how AI is beating scammers at their own game

In a groundbreaking initiative, British mobile operator O2 has developed a conversational AI named Daisy, designed to engage with scammers and effectively waste their time. Unlike traditional methods that block or filter scam calls, Daisy engages with fraudsters in a warm and chatty manner, keeping them occupied for extended periods, sometimes up to 40 minutes. This innovative approach not only disrupts the scammers' operations but also highlights the psychological tactics that scammers employ to manipulate their victims.

Daisy’s creation stems from a desire to combat the growing threat of phone scams, which often target vulnerable individuals, particularly the elderly. Scammers rely heavily on psychological manipulation, employing urgency and authority to bypass rational thought and instill fear in their targets. The effectiveness of these scams is not due to the sophistication of the targets but rather the psychological pressure exerted by the scammers. Daisy, with her friendly demeanor and scatterbrained persona, serves as a countermeasure that turns the tables on these con artists by engaging them in a manner that undermines their tactics.

The Mechanics of Daisy

Daisy operates autonomously, utilizing a dedicated phone number that has been strategically placed on “mugs lists,” which scammers use to identify potential victims. This clever infiltration allows Daisy to appear as the ideal target for fraudsters, thereby luring them into a trap. By embodying the characteristics that scammers typically exploit—such as being an older woman who seems unsure about technology—Daisy effectively disrupts the scammers' confidence and manipulative scripts.

The design of Daisy is not just a technical achievement; it is a psychological strategy that exploits the biases inherent in the scammers' profiling methods. By engaging with them in a non-compliant manner, Daisy keeps them on the line, thereby preventing them from targeting actual victims. This approach is reminiscent of the long-standing practice of scambaiting, where individuals would waste scammers' time with pre-recorded messages. However, Daisy takes this concept to a new level by operating independently and continuously, providing a scalable solution to the problem of phone scams.

Broader Implications for Technology and Fraud Prevention

The introduction of Daisy reflects a broader trend in technology where AI is being utilized not just for efficiency but also for protective measures against fraud. Similar systems have been implemented in mobile operating systems, such as Apple's iOS 26 and Google's Android, which feature call screening and real-time scam detection. These systems aim to introduce a moment of pause in the conversation, allowing potential victims to regain their composure and make informed decisions about whether to engage with the caller. This design principle emphasizes the importance of creating friction in the scammer's approach to prevent immediate compliance from targets.

Moreover, financial institutions are adapting to the evolving landscape of fraud with AI-driven solutions that analyze transaction patterns in real time. For instance, Mastercard's Consumer Fraud Risk system assesses transactions and the conversations leading up to them, marking a shift from static fraud detection to adaptive systems that learn individual behaviors. This evolution in fraud detection signifies a recognition that the psychological pressure exerted on victims is a critical factor in successful scams.

However, while these technological advancements are promising, they do not address the root causes of the problem. Many scammers are victims themselves, coerced into fraudulent activities under dire circumstances, such as modern slavery. This reality complicates the narrative around fraud and highlights the need for a multifaceted approach that encompasses governance, labor rights, and accountability in the tech industry.

The fight against scams is not merely a user experience issue but a complex challenge that intersects with organized crime and systemic exploitation. As governments and regulatory bodies begin to implement measures aimed at curbing fraud, the tech industry must also take responsibility for the infrastructure that enables these crimes. The Online Safety Act and similar legislation represent steps in the right direction, but the enforcement of these laws must keep pace with their ambitions.

In conclusion, Daisy serves as a proof of concept for the potential of AI to engage with and disrupt fraudulent activities. The design interventions that introduce pauses and friction into the scam process are not only effective but also necessary in a landscape where psychological manipulation is prevalent. However, the broader implications of fraud require a concerted effort from various stakeholders, including technology companies, governments, and society at large, to address the systemic issues that underpin these criminal operations.

As the industry continues to innovate in the realm of fraud prevention, it is crucial to remember that technology alone cannot solve the problem. A comprehensive approach that combines technological solutions with ethical considerations and social responsibility is essential to combat the complex issue of fraud at scale.