One of the main concerns regarding faster payments is the increased risk of fraud, but artificial intelligence (AI) can help address these issues.

A research study by the Bank for International Settlements (BIS) and the Bank of England evaluated AI’s capability to detect sophisticated fraudulent activities conducted by cybercriminals.

The experiments were based on data from millions of bank accounts and transactions, simulating real-time retail payments environments.

Project Hertha, as the study was called, revealed that AI models are highly effective fraud detection tools. According to BIS reports, AI was 26% more efficient at identifying suspicious activities compared to traditional fraud prevention methods.

In addition, AI analytics assisted financial institutions in uncovering fraudulent accounts that they would not have identified otherwise, increasing their discoveries by 12%.

Powerful Evolution

The potency of AI in fraud protection was further highlighted by data from FIS. Among the respondents, 78% reported that AI has improved their company’s fraud detection and risk management strategies. Nearly half of these leaders plan to increase their investment in AI over the next two years, with many intending to automate more complex tasks.

One significant development is agentic AI, where AI agents can handle numerous tasks autonomously. While agentic AI has the potential to combat fraud effectively, many experts view it as a double-edged sword.

A research study by SailPoint found that 96% of tech professionals consider AI agents a growing security threat. However, nearly all respondents expect to expand their use of agentic AI in the coming year.

Supplement, Not a Solution

As organizations adopt AI technologies, cybercriminals have already deployed both generative and agentic AI on a large scale, utilizing these tools for various fraudulent activities such as deepfakes and ransomware attacks. One reason why cybercriminals have gained significant advantages is their lack of concern over privacy or reputation issues.

Despite Project Hertha’s findings, there are limitations to AI models that can either miss instances of fraud or generate false positives. This led BIS to conclude that while AI should be seen as a valuable supplement to existing fraud defenses, it cannot serve as the sole solution. Organizations must continue to innovate and develop new strategies to keep pace with cybercriminals who have a substantial head start.