MAS launches AI risk toolkit for financial institutions with case studies from DBS, peers

[SINGAPORE] Financial institutions seeking guidance on implementing artificial intelligence (AI) risk management frameworks can now turn to a new set of handbooks, which includes case studies from leading industry players.

The publications form part of an AI Risk Management Toolkit released following the completion of the second phase of Project MindForge, an industry initiative led by the Monetary Authority of Singapore (MAS).

“This toolkit, developed collaboratively by a consortium of 24 leading banks, insurance companies, capital market firms and other industry partners, provides financial institutions with resources for managing AI-related risks across traditional AI, generative AI and emerging agentic AI technologies,” said MAS in a statement on Friday (Mar 20).

Participants include DBS, Julius Baer, OCBC, Standard Chartered and UOB, as well as insurers Income Insurance and Prudential, and asset manager BlackRock.

Launched in mid-2023, the first phase of Project MindForge focused on developing a generative AI risk framework to support responsible adoption by banks. It led to a white paper titled Emerging Risks and Opportunities of Generative AI for Banks.

The second phase broadened the scope to cover insurance and asset management, and expanded beyond generative AI to address a wider range of technologies, including emerging agentic systems.

The resulting toolkit comprises three documents: a 26-page executive handbook outlining strategic considerations and implementation practices; a 173-page operational handbook with detailed guidance and examples; and a 23-page publication featuring implementation case studies.

The case studies feature risk management approaches by DBS, Julius Baer, Prudential and an unnamed investment firm.

“At DBS, we recognise that the responsible use of data is increasingly essential as we expand the scale and pervasiveness of AI across the bank,” the lender wrote. “As we continue to explore the potential of AI, we remain steadfastly cautious of its associated risks.”

Within the bank, risk management is anchored by a Responsible Data Use framework, which evaluates AI use cases across legal and security requirements, ethical considerations and governance controls over deployment.

This is complemented by a risk-based AI governance approach, where use cases are assessed for materiality and subject to oversight, with accountability extending to senior management and the board.

Additionally, all AI models are tracked in a central repository, enabling end-to-end monitoring across their life cycle.

One example is CodeBuddy, an in-house generative AI coding assistant designed to boost productivity while incorporating safeguards such as human oversight, strict data controls and phased deployment.

To support wider industry adoption, MAS will also set up an AI risk management workgroup comprising MindForge participants and other practitioners under its BuildFin.ai initiative.

The initiative brings together technology providers, research institutes and financial institutions to develop implementation resources, facilitate knowledge sharing and build capabilities to manage risks from newer technologies such as agentic AI.

“The development of the MindForge AI Risk Management Toolkit… marks a major step forward in our journey to ensure the responsible adoption of AI in finance,” said MAS’ chief fintech officer Kenneth Gay.

“We are committed to fostering a culture of continuous engagement and strengthening of AI governance and risk management practices across the industry,” he added.