AI-driven fraud detection systems: A comprehensive review of emerging techniques in U.S. financial crime prevention

Financial fraud remains an enormous threat to the health of the contemporary financial system, especially with the fast growth of digital banking, internet-based payment systems, and fintech applications. The existing rule-based and statistical-based fraud detection systems have not been efficient in detecting sophisticated and dynamic fraud schemes in extremely interconnected financial ecosystems. In response, artificial intelligence (AI) has become a strong technology that can enhance the ability to detect fraud cases in financial intermediaries. This study gives a critical overview of the new AI-based methods in detecting financial fraud and their application in preventing financial crimes in the U.S. The study synthesizes current applications of AI across banking operations, anti-money laundering (AML) frameworks, and fintech payment ecosystems while evaluating commonly used datasets and performance metrics for fraud detection research. The analysis further reveals the main issues that can impact AI-based systems of detecting fraud, such as an imbalance between classes in the fraud data, attacker (or adversarial) attacks, insufficient data, interpretability of models, and privacy concerns. In addition, the latest research topics like explainable artificial intelligence, graph neural networks, federated learning, and real-time fraud detection systems are reported as potential ways to enhance the effectiveness of fraud detection. As a review of recent findings in AI-based fraud detection studies, this article offers important information to researchers, financial institutions, and policymakers in the field by ensuring the enhancement of financial crime prevention strategies and the resilience of digital financial systems.