Singapore's financial regulator has put a date on AI governance. On 7 October 2026, the Monetary Authority of Singapore (MAS) published its Guidelines on Artificial Intelligence Risk Management, setting supervisory expectations for financial institutions using AI — from traditional machine-learning models to Generative AI and increasingly autonomous AI agents. The guidelines take effect on 7 October 2027, giving regulated firms roughly one year to build the core governance layer and another year to complete the broader operational framework.

The document follows a public consultation launched in November 2025. Its scope is deliberately wide: MAS says the guidelines apply to all financial institutions and all forms of AI technology.

What does the framework actually mandate?

MAS has drawn a regulatory line, not a prohibition. The central principle is proportionality. Institutions must understand where AI is deployed, assess how important and risky each application is, and apply controls matching that risk. A tool proofreading an internal email does not demand the same controls as a model influencing investment recommendations or customer risk assessments.

The status of the document matters for compliance planning. These are guidelines setting out MAS' supervisory expectations, not standalone legislation imposing identical technical requirements on every firm. But firms should not read that as optional. MAS states that all financial institutions should adhere to its high-level expectations when deploying AI, with implementation proportionate to their size, activities and AI risk exposure. The guidelines complement existing MAS requirements and principles, including Fairness, Ethics, Accountability and Transparency — the FEAT framework.

Which firms are in scope?

The categories MAS lists are broad:

• Banks
• Fund managers
• Financial advisers
• Insurers
• Payment institutions
• Exchanges and clearing houses
• Custodians
• Credit rating agencies
• Firms dealing in capital markets products
• Exempt futures and OTC derivatives brokers, explicitly included among the covered categories

That last point deserves attention from the dealing desk. Brokers operating in futures and OTC derivatives under exemptions — often outside the perimeter of other MAS rulebooks — are named directly, which pulls their AI tooling into supervisory scope.

What risks does MAS identify?

The regulator's risk taxonomy goes well beyond model error. MAS specifically warns that AI agents with access to tools could autonomously perform erroneous or unauthorised actions — a direct operational and compliance concern for any desk experimenting with agentic workflows in execution, order handling or client interaction.

The guidelines identify financial, operational, conduct, financial-crime and reputational risks arising from AI, alongside security, privacy and third-party risks associated particularly with Generative AI. For institutions using vendor-hosted large language models or third-party AI infrastructure, the third-party risk dimension signals that outsourcing the model does not outsource accountability.

Why does this matter for execution workflows?

AI is moving beyond chatbots and productivity tools into areas that influence financial decisions, customer interactions, compliance, risk management and potentially execution itself, according to the regulator's own framing. Firms that have deployed machine learning for trade surveillance, client risk scoring or recommendation engines will need to document where those models sit, rate their materiality, and evidence controls proportionate to that rating — before the 7 October 2027 effective date.

The timeline creates a two-phase practical problem: governance inventory and risk classification first, then the operational control layer. Institutions with distributed AI adoption across front, middle and back office face the heavier lift, because the framework presumes an institution-wide view of AI usage.

MAS has also said it intends to consult the industry in 2027 on whether further guidance specifically addressing agentic AI would be useful — a signal that the current document may not be the last word on autonomous systems in finance.

For now, the mandated action is clear: firms have until October 2027 to stand up proportionate AI risk management, and MAS will be watching whether the proportionality defence holds up against the inventory of models actually running in production.