AI Regulation Tracker / Banking and finance

MAS Proposes AI Risk Management Guidelines for Every Financial Institution

Consultation issued November 13, 2025 and closed January 31, 2026. Singapore's MAS put out draft AI Risk Management Guidelines that would set supervisory expectations for all financial institutions, with a proposed 12-month transition. They are guidelines, not law, and they explicitly reach generative AI and AI agents.

By Anthony Guerriero · The Leveraged Years Regulatory Research · Published July 21, 2026 · Last updated July 21, 2026

The short version

The Monetary Authority of Singapore issued a consultation paper on November 13, 2025 proposing Guidelines on Artificial Intelligence Risk Management for financial institutions. The consultation closed on January 31, 2026. These are guidelines, which in MAS practice are supervisory expectations rather than binding legislation, and MAS proposes a 12-month transition period after they are issued for institutions to implement them. What makes them worth reading now is scope and breadth. They would apply to all financial institutions, including banks, insurers, capital markets firms, and financial advisers, and they cover the full AI life cycle from board and senior management oversight through data management, fairness, transparency and explainability, and human oversight. They explicitly reach generative AI and autonomous AI agents. For advisers and finance firms using AI, this is the shape of what MAS will expect, and the smart move is to benchmark your governance against the draft now rather than wait for the final text and a shrinking transition window.

Regulatory Briefing

Instrument: Consultation Paper on Guidelines on Artificial Intelligence Risk Management (AIRG)
Authority: Monetary Authority of Singapore (MAS)
Jurisdiction: Singapore; all MAS-supervised financial institutions
Bindingness: Non-binding. A consultation on proposed guidelines, which set supervisory expectations rather than statutory obligations. Not in force; still being finalized after the consultation closed.
What Changed: MAS moved from principles and thematic reviews toward a single, sector-wide set of supervisory expectations for AI risk management, covering the full life cycle and naming generative AI and AI agents.
Who This Affects: Financial advisers, banks, insurers, capital markets firms, and their boards, risk, compliance, and technology functions operating in or serving Singapore.
Current Status: Consultation issued November 13, 2025; closed January 31, 2026. Guidelines being finalized. A 12-month transition is proposed after they are issued.
Primary Source: MAS consultation page, mas.gov.sg
Last Verified: July 21, 2026

Singapore has spent years signaling on AI in finance through principles and thematic work, but always in pieces. This consultation is the moment MAS pulls it into one place. Issued on November 13, 2025 and closed to comment on January 31, 2026, the proposed Guidelines on Artificial Intelligence Risk Management would set MAS supervisory expectations for how financial institutions govern, test, monitor, and manage the risks of using AI. The reach is deliberately wide. MAS proposes to apply them across all financial institutions and to have them applied proportionately to the size and nature of each firm's activities and risk profile.

What would the MAS guidelines actually require?

They set expectations across the whole AI life cycle rather than a single control. MAS puts governance at the top. In the draft, "the board and senior management of FIs play a key role in the governance and oversight of AI risk management," which is MAS making clear that AI risk is a leadership responsibility, not something buried in the technology team. From there the guidelines reach into "robust controls in areas such as data management, fairness, transparency and explainability, human oversight," along with the systems, policies, procedures, and capabilities a firm needs to run AI responsibly. Critically, the draft names the newer technology directly. It covers generative AI and autonomous AI agents, not only traditional models, which means the tools most firms are rushing to deploy are squarely in view.

Are the MAS AI guidelines mandatory?

Not as drafted, and it is worth being precise. These are guidelines, and in MAS practice guidelines set out supervisory expectations rather than hard statutory obligations. They are also still a proposal. The consultation closed on January 31, 2026, and MAS is working through the responses before it issues final guidelines. So nothing here is in force today, and even once issued the instrument will be a set of expectations rather than a rule with direct penalties. That said, treating MAS supervisory expectations as optional would be naive. MAS supervises against its guidelines, and a firm that cannot show it meets them will feel that in supervisory engagement well before any formal action.

What should financial advisers and finance firms do now?

Use the transition math to your advantage. MAS proposes a 12-month implementation window after the guidelines are issued, which sounds generous until you remember that building AI governance across data, fairness, explainability, human oversight, and board reporting is a program, not a policy memo. The firms that wait for the final text will start that clock from zero. The firms that read the draft now can spend the interim mapping their current AI use against the proposed expectations, finding where they lack ownership, monitoring, or explainability, and closing those gaps before the window even opens. If you use or plan to use generative AI or AI agents in advice, underwriting, or client servicing, assume they are in scope, document how a human stays accountable for the output, and be able to explain to a supervisor how the tool is governed. None of that is wasted work even if the final guidelines shift at the margins, because the direction of travel is unambiguous.

Questions professionals are asking

Are the MAS AI Risk Management Guidelines in force?

No. They are a proposal. MAS issued the consultation on November 13, 2025 and it closed on January 31, 2026. MAS is finalizing the guidelines. Even once issued, they are supervisory expectations rather than binding legislation, and a 12-month transition is proposed before firms are expected to have implemented them.

Which firms would they apply to?

All MAS-supervised financial institutions, including banks, insurers, capital markets firms, and financial advisers. MAS proposes to apply them proportionately, commensurate with the size and nature of each firm's activities, use of AI, and risk profile.

Do the guidelines cover generative AI and AI agents?

Yes. The draft expressly covers generative AI and autonomous AI agents alongside traditional models. It sets expectations across the AI life cycle, including data management, fairness, transparency and explainability, and human oversight, with the board and senior management responsible for governance.

What should a financial adviser do before the guidelines are final?

Benchmark now. Map your current and planned AI use against the proposed expectations, identify gaps in ownership, monitoring, and explainability, and document how a human stays accountable for AI-assisted advice or decisions. Starting during the consultation-to-issuance window leaves you far better placed than firms that wait and start the 12-month clock from zero.

RELATED BRIEFINGS

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any guideline or requirement applies to your firm with qualified counsel in the relevant jurisdiction.