Singapore Model AI Governance Framework for Agentic AI
Issued by
Infocomm Media Development Authority (IMDA), Singapore
- September 30, 2026 · Correction — Rewrote this entry around the framework itself (launched 22 January 2026); it previously described only a later parliamentary answer. Checked against IMDA and MDDI. (Cody Maxwell)
- September 30, 2026 · Correction — Corrected the update date (version 1.5, 20 May 2026) and rewrote the FAQ, which wrongly described mandatory duties and licensing consequences (fact-check finding). (Cody Maxwell)
- October 1, 2026 · Correction — Rewrote the practical steps, newsletter hook, search description, and audience fields to match the entry's corrected content. (Cody Maxwell)
- October 2, 2026 · Correction — Removed an unsupported claim that AI Verify will be extended to agentic systems; the August 2026 parliamentary replies announced no such step. (Cody Maxwell)
Singapore's IMDA launched the Model AI Governance Framework for Agentic AI in January 2026. It is the first government framework written for AI agents that plan and act on their own. It is voluntary, but organizations remain legally accountable for their agents' actions.
Applies To
Overview
Singapore's Infocomm Media Development Authority (IMDA) launched this framework on 22 January 2026 at the World Economic Forum. It extends Singapore's Model AI Governance Framework to AI agents that can plan, reason, and take actions with limited human direction. The framework is organized around four areas. Organizations should assess and bound the risks of each agent, keep meaningful human accountability, apply technical controls and processes, and help end users act responsibly. It applies whether an organization builds its own agents or uses third-party ones. Following the framework is voluntary, but organizations remain legally accountable for what their agents do. IMDA published version 1.5 on 20 May 2026, adding guidance on multi-agent systems, third-party agents, and case studies. Asked in Parliament in August 2026 whether AI Verify would cover agentic systems, the Ministry of Digital Development and Information pointed to this framework and announced no new measures.
Key Requirements
- •Assess each agent's risks and set bounds on what it can access and do.
- •Keep meaningful human accountability, with named owners and checkpoints for consequential actions.
- •Apply technical controls and processes across the agent's life cycle, including testing and monitoring.
- •Give end users the information they need to use agents responsibly.
- •Apply the framework to third-party agents as well as agents built in-house.
What Your Organization Must Do
- →Map every AI agent your organization runs, including agents supplied by third-party vendors.
- →Set written bounds on what each agent can access and do.
- →Name an accountable owner for each agent and define human checkpoints before consequential actions.
- →Test and monitor agents across their life cycle, and keep records of the results.
- →Give end users clear information on how to use agents responsibly.
- →Confirm existing legal duties, such as the Personal Data Protection Act, still cover agent actions.
Playbook Guidance
Step-by-step implementation guidance for compliance teams.
Governance Controls
Operational controls that implement requirements from this regulation.
Frequently Asked Questions
- Is Singapore's Model AI Governance Framework for Agentic AI legally binding?
- No, it is voluntary guidance with no penalties, registration, or enforcement of its own. Organizations remain legally accountable for their agents' actions under existing laws, such as Singapore's Personal Data Protection Act.
- Which organizations does the framework apply to?
- It is written for any organization deploying agentic AI in Singapore. That includes organizations that build their own agents and those that use agents from third parties.
- What counts as an agentic AI system under the framework?
- The framework covers AI systems that can plan, reason, and take actions toward a goal with limited human direction. It treats them as needing governance beyond that of conventional software.
- What does the framework recommend on human oversight of AI agents?
- It recommends meaningful human accountability, such as named owners for agents and human checkpoints before consequential actions. These are recommendations, not legal requirements.
- How does AI Verify relate to the agentic framework?
- AI Verify is Singapore's separate testing toolkit for AI systems. Asked in August 2026 whether it would be extended to agentic systems, the Ministry of Digital Development and Information pointed to this framework and announced no extension.
- What happens if a company does not follow the framework?
- The framework itself carries no penalties or sanctions. The practical exposure comes from existing laws and sector rules, which still apply to whatever an agent does.