Explainable and Multimodal AI for Financial Fraud Prevention: Advancing Trust and Cybersecurity in Digital Banking

Digital banking has made financial services faster, more accessible, and cheaper to run. It has also widened the opening for financial fraud, identity theft, social engineering, synthetic identity attacks, and deepfake-enabled cybercrime. Rule-based security systems often struggle with threats that are sophisticated and change quickly, which has increased interest in adaptive artificial intelligence (AI) solutions. This review examines how explainable artificial intelligence (XAI) and multimodal AI can strengthen fraud prevention, support decision-making, and improve cybersecurity resilience in digital banking. It looks at the integration of multiple data modalities, real-time threat detection, explainable fraud alerts, and organizational trust in AI-generated decisions. The review synthesizes two empirical studies, one on U.S. digital banking and one on the U.S. banking sector. The first reports that multimodal deepfake detection, real-time detection, and explainable AI are each positively associated with fraud detection effectiveness and cybersecurity resilience. The second reports that explainable AI trust is a significant link between AI-driven fraud detection and banking cybersecurity resilience, while its proposed real-time fraud response mediation was not statistically supported. The review also covers methodological limitations and challenges involving data privacy, model transparency, false positives, adversarial attacks, and organizational readiness. It concludes that AI-based financial security needs interpretable decisions, human oversight, timely operational action, and responsible governance in addition to accurate detection.