The Monetary Authority of Singapore concluded Project MindForge Phase 2 in early 2026, publishing an AI Risk Management Operationalisation Handbook. The handbook was developed by a consortium of 24 banks, insurers, and capital market firms under MAS facilitation. It is a published guidance instrument that MAS-regulated financial institutions may use to operationalise AI risk management in compliance with MAS supervisory expectations.
The handbook applies alongside the MAS Guidelines for Artificial Intelligence Risk Management, issued under MAS's supervisory authority over financial institutions licensed under the Monetary Authority of Singapore Act, the Banking Act, the Insurance Act, and the Securities and Futures Act. The MAS Guidelines set out supervisory expectations across four dimensions: board and senior management oversight of AI, AI life cycle controls, risk management systems and policies, and required institutional capabilities. The operationalisation handbook translates those expectations into tool-level implementation guidance across three AI paradigms: traditional AI models, generative AI models, and agentic AI systems.
Banks, insurers, capital market intermediaries, and payment institution licensees operating in Singapore face direct supervisory exposure under the MAS AI risk program. Firms using AI for credit decisions, fraud detection, customer-facing recommendation services, agentic trading or advisory functions, and internal compliance operations face the highest MAS scrutiny. Institutions with material exposures to generative AI in client-facing or compliance roles will find the handbook's treatment of hallucination risk, model governance, and human oversight requirements of immediate practical relevance.
The MAS program applies proportionately to each institution based on the scale and complexity of its AI use. MAS has not published a mandatory implementation deadline for full compliance with the handbook. The final AI Risk Management Guidelines remain principles-based rather than prescriptive on specific technical standards. An open question is how the program interacts with third-party AI models procured from external providers, given that MAS supervisory expectations on model governance apply regardless of whether a model is developed in-house or externally sourced.
Licentium advises financial institutions, fintech firms, and AI developers on MAS regulatory compliance, AI governance programs, and Singapore licensing strategy. We may assist directly or through our partner network specialising in MAS-regulated financial services. Work we undertake includes AI risk management program design, MAS examination preparation, fintech licensing, agentic AI compliance programs, and cross-border AI regulatory strategy across Asia-Pacific financial regulators.