CBA Research Reveals Human Vulnerability in AI Scam Detection

• •90% of Australians overestimate their ability to detect AI scams, with only 42% accurate.
• •Human decision-making is the main vulnerability in scam prevention, often leading to financial losses.
• •CBA's AI agents have conducted over 350,000 interactions to gather intelligence against scammers.

Commonwealth Bank's recent research indicates a significant gap between Australians' confidence in identifying AI-generated deepfake scams and their actual detection capabilities, with nearly 90% believing they could spot such scams but only 42% succeeding in tests. The study highlights that human decision-making remains the primary vulnerability, as one in three scam victims suspected fraud yet proceeded with payments due to urgency. The research also emphasizes that deception occurs long before payment is made, with scammers building credibility through impersonation of trusted entities. In 2025, payment redirection scams cost Australians $166.8 million, showcasing the financial impact of these scams. To combat these threats, CBA has deployed AI agents that have engaged in over 350,000 interactions with scammers, feeding intelligence back into fraud detection systems. Collaborative initiatives like the Scam-Safe Accord are also being implemented to enhance coordination across the payments ecosystem.