AI Boosts Retail Payment Fraud Detection by 12%
In an era where digital payments have become the lifeblood of global commerce, the fight against payment fraud is more critical than ever. Fraudsters are getting smarter, and so must the systems designed to stop them. Enter Project Hertha—a groundbreaking collaboration between the Bank for International Settlements (BIS) Innovation Hub and the Bank of England (BoE)—that’s pushing the boundaries of artificial intelligence to detect fraud in retail payments in real time. As of June 2025, this initiative is not just a proof of concept but a beacon showing how AI can transform financial crime detection at the systemic level.
Breaking New Ground: Why AI for Retail Payment Fraud?
Let’s face it: traditional fraud detection methods are often reactive, relying heavily on historical data and manual intervention. With retail payment volumes skyrocketing—running into billions of transactions daily worldwide—the challenge is to spot suspicious activities as they unfold. The costs are staggering; payment fraud losses globally hit over $40 billion in 2024 alone, according to industry analysts. That’s where AI steps in, promising to analyze massive datasets at speeds and depths no human team can match, identifying subtle, complex patterns that hint at coordinated criminal activity.
Project Hertha leverages "modern AI techniques," including advanced machine learning algorithms and network analysis tools, to detect fraud patterns across entire payment ecosystems instead of isolated accounts. This systemic approach is a paradigm shift, aiming to spot fraud rings and novel attack methods in real time—not days or weeks later.
How Project Hertha Works: A Six-Stage Feedback Loop
The project tested these AI models on a simulated dataset comprising 1.8 million synthetic bank accounts and more than 300 million transactions, mirroring real-world retail payment flows. The results were impressive: the AI detected 12% more fraudulent activities than traditional methods alone—a significant uplift that could translate to billions of dollars saved annually[4].
At the core of Hertha’s success is a sophisticated feedback loop between payment systems and payment service providers (PSPs):
- Detection: AI flags suspicious transaction patterns in real-time.
- Alert: The system sends these alerts to PSPs for further investigation.
- Investigation: PSPs analyze flagged cases, validating or dismissing alerts.
- Feedback: Results of investigations are fed back into the AI system.
- Learning: AI updates its models based on the feedback, improving accuracy.
- Repeat: This cycle continues, allowing the system to adapt to evolving fraud techniques.
This continuous learning process is key. Without it, the AI models risk becoming stale, generating too many false positives and missing new fraud tactics. The feedback loop ensures the system evolves with the threat landscape, maintaining its edge.
Transparency Matters: Explainable AI in Fraud Detection
One standout insight from Project Hertha is the importance of explainability. Banks and PSPs are understandably cautious about automated alerts—false alarms can waste resources, while opaque AI decisions breed mistrust. Hertha's models were designed to provide clear reasons for flagging suspicious behavior, detailing which transaction characteristics or network relationships triggered the alert and suggesting potential fraud types. This transparency encourages action and collaboration across institutions, which historically operated in silos.
As a BIS report explained, explainable AI helps bridge the gap between cutting-edge technology and practical, real-world deployment by fostering trust and operational clarity[3].
The Broader Implications for Financial Systems
Project Hertha is more than a fraud-fighting tool; it’s a step toward smarter, safer payment infrastructures. By enabling payment systems to do more than just move money, it transforms them into active guardians of financial integrity. The project also highlights the value of data sharing—when done responsibly and with privacy safeguards, combining insights across institutions amplifies fraud detection capabilities.
This initiative complements ongoing efforts worldwide to modernize payment security. For example, regulators in the EU and US are encouraging the adoption of AI-driven fraud detection frameworks, and private sector players like Mastercard and Visa have been investing heavily in AI-powered fraud analytics. But Hertha’s collaborative, cross-institutional approach sets a new standard for how central banks and financial institutions can jointly tackle systemic risks.
Challenges and the Road Ahead
Despite its promise, deploying AI on this scale isn’t without hurdles. Data privacy and security remain paramount—ensuring that sensitive payment information is protected while enabling meaningful analysis requires sophisticated governance frameworks. Moreover, integrating AI alerts into existing fraud investigation workflows demands training and cultural shifts within institutions.
Another challenge is maintaining model robustness against adversarial attacks—fraudsters are quick to adapt once they understand AI detection patterns. Continuous model retraining and incorporating diverse data sources will be essential to stay ahead.
Looking forward, Project Hertha is poised to expand beyond retail payments, potentially applying AI to wholesale payment systems and cross-border transactions. The BIS Innovation Hub plans to foster similar initiatives globally, creating a network of AI-enhanced financial crime detection capabilities that could redefine how the world safeguards its money.
Conclusion: AI’s Role in the Future of Payment Security
As someone who’s followed AI’s evolution in finance for years, Project Hertha feels like a pivotal moment. It encapsulates how artificial intelligence, when thoughtfully integrated with human expertise and institutional collaboration, can revolutionize fraud detection. By boosting detection rates by 12% and fostering explainability and continuous learning, this project lays a blueprint for safer, smarter payment ecosystems in an increasingly digital world.
The fight against payment fraud is a moving target, but with AI as an ally, the financial industry is better equipped than ever to protect consumers and institutions alike. If Project Hertha is any indication, the future of payment security will be digital, dynamic, and decidedly more resilient.
**