Enhancing Banking Authentication Systems with LLMs and RAG: Addressing Ethical and Privacy Challenges

Banking access control security can be improved by integrating Large Language Models (LLMs) customized with Retrieval-Augmented Generation (RAG) architecture into banking authentication systems. Deploying these systems may cause significant ethical challenges and privacy-related issues that expose data, produce biased decisions, and force companies to break the rules and regulations. This paper investigates these security issues using a Design Science Research (DSR)-based LLM and RAG enabled banking authentication system. The artifact incorporates privacy-preserving artificial intelligence techniques and explainable artificial intelligence functions to protect data integrity and meet General Data Protection Regulation (GDPR) and Payment Card Industry Data Security Standard (PCIDSS) regulatory criteria. An extensive set of experimental simulations involving 550 typical banking authentication use cases confirmed that security systems built upon LLM and RAG technology produce superior banking authentication outcomes regarding transparency, fairness, and privacy safeguards. This research indicates that AI-based banking authentication systems must focus on responsible AI governance, regulatory adherence, and fairness protocols to provide secure bank environments.