IMDA launched the Model AI Governance Framework for Agentic AI at the World Economic Forum on 22 January 2026 and updated it on 20 May 2026 (version 1.5) with feedback from over 60 organisations, including DBS, and more than ten case studies. It sets four dimensions for deploying AI agents: assess and bound the risks upfront, make humans meaningfully accountable, implement technical controls and processes, and enable end-user responsibility. It is voluntary, but MAS's Guidelines on AI Risk Management point financial institutions to it for agentic AI governance, and the update includes an OCBC case study bounding agent autonomy in source-of-wealth analysis.
OFFICIAL TEXT: imda.gov.sg ↗ · IN FORCE · IMDA SINGAPORE
| Document | IMDA Model AI Governance Framework for Agentic AI — Model AI Governance Framework for Agentic AI |
| Issued by | Infocomm Media Development Authority (Singapore), with the AI Verify Foundation |
| Type | Framework |
| Status | In force |
| Published | Jan 22, 2026 |
| Applies to | Organisations deploying agentic AI, 'whether through developing AI agents in-house or using third-party agentic solutions', including banks; voluntary IMDA guidance and a 'living document' |
| Official source | imda.gov.sg ↗ |
| Use cases | Generative & agentic AI · AI governance (general) · Third-party & vendor AI · AML / KYC |
What are the key points of IMDA Model AI Governance Framework for Agentic AI?
- Four dimensions: 'Assessing and bounding the risks upfront' (use-case selection and limits on autonomy and access to tools and data), 'Making humans meaningfully accountable' (significant checkpoints requiring human approval), 'Implementing technical controls and processes' through the agent lifecycle, and 'Enabling end-user responsibility'.
- IMDA's launch release says agents may be able to make changes such as 'updating a customer database or making a payment', creating risks of 'unauthorised or erroneous actions' and automation bias.
- Version 1.5 (20 May 2026) adds guidance on multi-agent systems, third-party agents and automation bias, based on feedback from more than 60 organisations such as AWS, DBS, Google and Salesforce.
- Case study: Bank of Singapore (OCBC) uses an agentic system to draft source-of-wealth memos; it 'provides decision support only and does not make credit, onboarding or risk decisions autonomously', with task-level autonomy, no self-initiated actions and human review at critical points.
- Other case studies include Ant International and a payroll agent whose transfers are capped by controls at the hardware level (Terminal 3).
- Footnote 11 of MAS's Guidelines on AI Risk Management names it as a reference for agentic AI risks and governance; the PDPC's 2026 generative AI guidelines and MAS's SAFR white paper also refer to it.
- IMDA says it is developing guidelines for testing agentic AI applications, building on its Starter Kit for testing LLM-based applications.
What did IMDA Model AI Governance Framework for Agentic AI change for banks?
This is the first government framework devoted to AI agents. It moves Singapore's governance guidance from model outputs to actions — bounding what agents can touch, defining where humans approve, and testing and monitoring agents in production — and gives banks a ready reference that MAS's own Guidelines endorse for agentic use cases.
Does MAS expect banks to follow IMDA's agentic AI framework?
MAS's Guidelines on AI Risk Management say institutions may reference frameworks such as IMDA's Model AI Governance Framework for Agentic AI for agentic-AI risks and governance. The IMDA framework itself is voluntary; MAS's Guidelines carry the supervisory expectations, and they cover AI agents.
What are the four dimensions of the Model AI Governance Framework for Agentic AI?
Assess and bound the risks upfront; make humans meaningfully accountable; implement technical controls and processes; and enable end-user responsibility.
When was the IMDA agentic AI framework last updated?
IMDA launched it on 22 January 2026 and published an updated version 1.5 on 20 May 2026 with case studies and new best practices on multi-agent systems, third-party agents and automation bias; the PDF notes a further update on 5 June 2026.
| Date | Document | Status |
|---|---|---|
| May 30, 2024 | IMDA Model AI Governance Framework for Generative AI — Model AI Governance Framework for Generative AI: Fostering a Trusted Ecosystem | In force |
| Jun 7, 2023 | AI Verify Testing Framework — AI Verify Testing Framework and Toolkit (AI Verify Foundation) | In force |
| Oct 15, 2026 | FSB AI sound practices (final report, Oct 2026) — Sound Practices for Responsible Adoption of Artificial Intelligence (AI): Final report | Final |
| Oct 15, 2026 | HKMA "ART" Framework / AI supervisory guidance — "ART" Framework and Updated Supervisory Guidance for Responsible Adoption of Artificial Intelligence | Final |
| Oct 7, 2026 | MAS Guidelines on Artificial Intelligence Risk Management — Guidelines on Artificial Intelligence Risk Management for Financial Institutions | Final |
| Sep 30, 2026 | Bailey: Frontier AI and the Question of Governance (Sep 2026) — Frontier AI and the Question of Governance — Governor Andrew Bailey | Final |
Follow every document these regulators publish
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning