Secure Prompt Engineering for Banking and Payment Applications Design Principles, Threat Models, and Governance Controls for Generative AI in Regulated Financial Systems

Abstract By December 2023, generative artificial intelligence systems based on large language models were increasingly evaluated for use in banking and payment applications, including customer support automation, internal operations assistance, fraud analysis support, compliance interpretation, and developer productivity tools. While these systems demonstrated strong natural language reasoning capabilities, their deployment within regulated financial environments introduced a new class of security and governance challenges centered on prompt construction, context control, and interaction boundaries. Unlike traditional software systems where inputs are tightly constrained and validated through deterministic logic, prompt driven systems rely on natural language instructions that directly influence model behavior, decision framing, and output generation.