MDDI Response on Extending AI Governance to Agentic AI Systems

Applies To

Overview

This parliamentary question response, published on 12 August 2026, provides official government clarification on the scope and application of Singapore's Model AI Governance Framework for Agentic AI. It confirms that agentic systems, defined by their capacity to plan and execute multi-step tasks with limited human intervention, fall within the framework's remit and must be governed accordingly. Key provisions address the designation of accountable roles for overseeing autonomous workflows, requirements for proportionate risk management controls, and the expectation that AI Verify testing tools will be extended to cover agentic system assessments. The response does not introduce new binding law but carries significant regulatory weight as an authoritative government interpretation of existing framework obligations. Enterprises operating in Singapore or deploying agentic AI systems that interact with Singapore residents should treat this response as clarifying compliance expectations under the existing framework. No formal penalty regime is attached, but non-alignment may affect regulatory standing and future licensing or procurement eligibility.

Key Requirements

• •Designate explicit oversight roles responsible for monitoring and intervening in agentic AI workflows, with documented accountability chains.
• •Implement risk management controls that are proportional to the level of autonomy granted to each agentic system.
• •Ensure meaningful human oversight is maintained at critical decision points, particularly where consequential actions can be initiated without direct human instruction.
• •Extend AI Verify assessment coverage to include agentic systems as the tool's scope is updated to support such evaluations.
• •Treat agentic AI deployments as distinct governance objects, separate from conventional software, with dedicated documentation and review cycles.
• •Align governance practices with the Model AI Governance Framework for Agentic AI as published in January 2026, incorporating this parliamentary clarification as authoritative interpretive guidance.

What Your Organization Must Do

• →Audit all agentic AI systems currently in production or under development and classify them by autonomy level and decision consequence.
• →Appoint named individuals as accountable oversight officers for each agentic workflow, and document their responsibilities formally in governance records.
• →Map each agentic system's decision points to determine where human checkpoints must be inserted or strengthened to satisfy meaningful oversight expectations.
• →Update internal AI governance policies to explicitly distinguish agentic systems from conventional software, and assign them to a dedicated review and approval process.
• →Review vendor and partner contracts involving agentic AI capabilities to confirm alignment with Singapore's framework requirements and this clarifying guidance.
• →Prepare for AI Verify scope expansion by cataloguing agentic systems now, so assessments can be initiated promptly once updated tooling is available.

Playbook Guidance

Step-by-step implementation guidance for compliance teams.

Frequently Asked Questions

Does Singapore's MDDI parliamentary response on agentic AI create new binding legal obligations for enterprises?

No new binding law is introduced, but the response carries significant regulatory weight as an authoritative government interpretation of the existing Model AI Governance Framework for Agentic AI. Non-alignment may affect regulatory standing and eligibility for future licensing or government procurement in Singapore.

Which organisations need to comply with Singapore's agentic AI governance framework clarified in the August 2026 parliamentary response?

The guidance applies to enterprises of all sizes, public sector bodies, AI developers, and AI deployers operating in Singapore or deploying agentic systems that interact with Singapore residents. This covers both local and international entities whose agentic AI workflows touch the Singapore market.

What counts as an agentic AI system under Singapore's Model AI Governance Framework as clarified by MDDI?

Agentic systems are defined by their capacity to plan and execute multi-step tasks with limited human intervention. The MDDI clarification confirms these systems must be governed as autonomous actors, distinct from conventional software features, and assessed under dedicated review processes.

What specific oversight roles must Singapore enterprises designate for agentic AI workflows under this framework?

Enterprises must appoint named individuals as accountable oversight officers for each agentic workflow, with documented accountability chains formalised in governance records. These roles carry responsibility for monitoring autonomous system behaviour and intervening at critical decision points where consequential actions can be initiated without direct human instruction.

How does AI Verify factor into compliance with Singapore's agentic AI governance requirements?

The MDDI response confirms that AI Verify testing tools will be extended to cover agentic system assessments as its scope is updated. Enterprises should catalogue their agentic systems now so that formal assessments can begin promptly once the updated tooling becomes available.

What penalties apply if a company's agentic AI systems are not aligned with Singapore's Model AI Governance Framework for Agentic AI?

No formal penalty regime is attached to this framework or the MDDI parliamentary clarification. However, non-alignment carries practical consequences including reduced regulatory standing and potential exclusion from government licensing processes and public sector procurement opportunities.