SG Agentic AI vs MAS AI Guidelines: what they share and how they differ

SG Agentic AI and MAS AI Guidelines share 12 of 24 control themes and 10 checks. 64% of MAS AI Guidelines requirements can be proven with checks SG Agentic AI already uses, 89% the other way round.

SGVoluntary frameworkVoluntary

SG Agentic AI

How to govern AI agents under Singapore's framework

9 requirementsNext 2027
SGSupervisory guidelinesUnder discussion

MAS AI Guidelines

Preparing for the MAS AI risk management guidelines

11 requirementsNext Dec 2026
12/24shared control themes
10shared checks
64%of MAS AI Guidelines requirements covered by SG Agentic AI evidence
89%of SG Agentic AI requirements covered by MAS AI Guidelines evidence

At a glance

SG Agentic AI MAS AI Guidelines
JurisdictionSingapore sameSingapore same
KindVoluntary frameworkSupervisory guidelines
StatusVoluntaryUnder discussion
BindingNo sameNo same
Object analysedAI system sameAI system same
ScopeAny organisation deploying AI agents in Singapore, built in-house or supplied by a third party.All financial institutions supervised by MAS (banks, insurers, asset managers, intermediaries).
Territorial reachSingapore; an international reference for agent governance.Financial activities in Singapore, including branches of foreign groups.
PenaltiesNone (voluntary). The organisation remains legally accountable for its agents' actions under existing law (PDPA, contract, tort).No penalty of their own. Once issued, they are supervisory expectations checked in inspections; shortcomings can lead to supervisory action.
Qualification axesAgent risk level (autonomy × access), Agent originUse-case risk materiality (impact, complexity, reliance)
RolesDeploying organisation, Agent developerFinancial institution
Requirements911
Next milestone2027, Follow-up to the discussion paper on agents' legal responsibilityDec 2026, Expected issuance of the final guidelines

Theme by theme

requirements per theme

What they share: one piece of evidence, two frameworks

10

CodeCheckRequirements SG Agentic AIRequirements MAS AI Guidelines
CHK-RISK-TOLERANCERisk tolerances are defined and AI systems are assigned to risk levels
CHK-CATEGORIZATIONAI system tasks and methods are categorized (classifier, generative, recommender)
VER-008-02System designed to allow human oversight (stop button, override)
CHK-INVENTORYA mechanism to inventory AI systems is in place and resourced
CHK-ROLES-CLARIFIEDRoles, responsibilities and delegated authorities are documented and clear to relevant stakeholders
VER-008-03Competent overseers assigned to the system
CHK-TEVVTEVV plan, test sets, metrics and data considerations are documented
VER-021-D-03Operation monitoring
VER-AUTO-01Suspension procedure in the event of risk
CHK-THIRDPARTY-POLPolicies address third-party AI/data risks, incl. IP, transparency and testing

Differences: requirements specific to each framework

Requirements with no check serving the other framework: the extra work.

Timelines

PastSet in the textPotentialTo verify
Nov 12, 2018MAS AI Guidelines · FEAT principles (fairness, ethics, accountability, transparency)
Jan 23, 2019SG Agentic AI · Model AI Governance Framework, 1st edition
Jan 21, 2020SG Agentic AI · Model AI Governance Framework, 2nd edition
May 25, 2022SG Agentic AI · AI Verify launched (governance testing)
May 30, 2024SG Agentic AI · Model AI Governance Framework for Generative AI
Dec 2024MAS AI Guidelines · Information paper on AI model risk management
Nov 13, 2025MAS AI Guidelines · Public consultation on the guidelines
Jan 22, 2026SG Agentic AI · Agentic AI framework published (WEF)
Jan 31, 2026MAS AI Guidelines · Consultation closes
May 20, 2026SG Agentic AI · Update: multi-agent systems, third-party agents, automation bias; discussion paper on legal responsibility for agents
Dec 2026MAS AI Guidelines · Expected issuance of the final guidelines
2027SG Agentic AI · Follow-up to the discussion paper on agents' legal responsibility
Dec 2027MAS AI Guidelines · End of the 12-month transition (if issued late 2026)

Run these requirements across all your AI systems

TrustFlow inventories your systems, qualifies them under each regulation and collects evidence once for every referential.

Request a demo