SG
Singapore | Regulatory Framework Status: in force
Effective: 2026-01-22
moderate

Singapore IMDA: Model AI Governance Framework for Agentic AI (v1.0)

Model AI Governance Framework for Agentic AI (Version 1.0)

I. Regulatory Summary

Provides a reference baseline for Singapore-facing legal and advisory teams to structure agentic AI governance programs (policies, risk assessment, access controls, oversight and monitoring), and to benchmark client practices against regulator-endorsed expectations.

II. Full Description

IMDA’s Model AI Governance Framework for Agentic AI (v1.0) describes agentic AI risks and emerging best practices for organisations deploying agents in-house or through third parties. It organises recommendations into four dimensions: (1) assessing and bounding risks via design limits and permissions; (2) ensuring human accountability and meaningful oversight; (3) lifecycle technical controls (guardrails, testing, monitoring); and (4) enabling end-user responsibility through information provision and training.

III. Scope & Application

Non-binding official governance framework for organisations developing or deploying agentic AI, providing practical governance expectations across four dimensions (risk bounding, human accountability, technical controls, and end-user responsibility).

IV. Policy Impact Assessment

Provides a reference baseline for Singapore-facing legal and advisory teams to structure agentic AI governance programs (policies, risk assessment, access controls, oversight and monitoring), and to benchmark client practices against regulator-endorsed expectations.

Primary Focus: AI governance framework (agentic AI)