The Role of Artificial Intelligence in Fraud Detection

The primary concern with faster payment systems is the potential for increased fraud. However, artificial intelligence (AI) can help mitigate these risks.

A study from the Bank for International Settlements (BIS) and the Bank of England explored AI’s effectiveness in identifying sophisticated fraudulent activities. The research was based on a simulation using data from millions of bank accounts and transactions, designed to mimic real-time retail payments.

The project, known as Project Hertha, revealed that AI models are highly effective fraud detection tools. According to the BIS report, AI was 26% more efficient than traditional methods in detecting suspicious activities.

Moreover, AI analytics enabled financial institutions to uncover an additional 12% of fraudulent accounts beyond what would have been identified otherwise.

The Evolution of Artificial Intelligence

The potency of AI in fraud prevention has also been highlighted by data from FIS. According to the findings, 78% of respondents reported that artificial intelligence had enhanced their company’s fraud detection and risk management strategies. Nearly half of these business and tech leaders plan to increase their investment in AI over the next two years, delegating more complex tasks to it.

A notable advancement is agentic AI, where AI agents can perform various tasks independently. While this technology offers significant advantages against fraud, many experts view it as a double-edged sword.

Security Threats and Challenges

Despite the benefits, research from SailPoint indicates that 96% of tech professionals consider AI agents to be a growing security threat. Nevertheless, nearly all respondents intend to expand their use of agentic AI in coming years.

Cybercriminals have already begun utilizing both generative and agentic AI on a large scale, employing them in diverse fraud schemes including deepfakes and ransomware attacks. One key reason cybercriminals are gaining an advantage is that they are not constrained by privacy or reputational concerns.

Although Project Hertha demonstrates the power of AI in detection, there remain limitations such as potential oversight errors—both missing genuine instances of fraud and producing false positives. These limitations led BIS to conclude that AI tools should be seen as a supplementary measure rather than a complete solution for fraud prevention.

Innovation and Out-of-the-Box Thought

To stay ahead of cybercriminals, organizations must think creatively. While AI is a valuable tool, they cannot rely solely on it. Instead, companies need to innovate new approaches to counteract the head start given by criminals.