Generative AI Guardrails in Banking

This Whitepaper outlines the framework for implementing guardrails in the banking sector to manage the risks associated with Generative AI (Gen AI). The document is a collaborative effort by the Association of Banks in Singapore (ABS) Standing Committee on Data Management (SCDM), which identifies specific risks that may arise from Gen AI usage and proposes corresponding guardrails. The approach is based on the collective experience of ABS SCDM members who have implemented Gen AI across various use cases. The Whitepaper details seven common enterprise Gen AI applications and the prevalent risks linked to them, along with strategies for mitigation. It emphasizes the importance of governance, training, and risk management practices. Additionally, practical considerations are illustrated through capability studies on Document Extraction/Summarisation and Code Generation, highlighting key risks and considerations for real-world implementation. This document aims to evolve with ongoing developments in Gen AI and will contribute to the AI Governance Handbook as part of Project MindForge.