Every AI Policy Ever · open in the interactive register

Published (non-binding, current)Non-binding · Guidelines / framework

Model AI Governance Framework for Agentic AI

Record SG-08

Issued by
Infocomm Media Development Authority (IMDA)
Jurisdiction
Singapore
Level · region
National · South & Southeast Asia
Published / adopted
2026-01-22
Last updated
2026-06-05
Language
English

Summary

Governance framework for organisations deploying AI agents, whether developed in-house or bought from third parties. It covers four dimensions: assess and bound the risks upfront by choosing suitable use cases and limiting agents' autonomy and access to tools and data; make humans meaningfully accountable, with checkpoints that require human approval and measures against automation bias; implement technical controls and processes across the agent lifecycle, such as access controls, baseline testing and continuous monitoring; and enable end-user responsibility through transparency and training.

Safety topics

human oversightaccountability & liabilityevaluation & testingtransparency & labelling

Status notes

Minister Josephine Teo announced Version 1.0 at the World Economic Forum in Davos on 22 Jan 2026; the framework builds on the governance foundations of the Model AI Governance Framework introduced in 2020. Version 1.5, published on 20 May 2026 and updated on 5 June 2026, incorporates feedback from more than 60 companies: it adds systemic and multi-agent risks, new risk factors such as system complexity and the use of third-party solutions, more practices against automation bias, an overview of types of controls, change-management recommendations and case studies, including IMDA's own on OpenClaw deployments. IMDA is also developing guidelines for testing agentic AI applications.

Sources

Open the official source https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/artificial-intelligence/mgf-for-agentic-ai.pdf

Secondary source: https://www.mddi.gov.sg/newsroom/singapore-launches-new-model-ai-governance-framework-for-agentic-ai--/

Local copy

Archived. Last attempt 2026-10-18 06:48:33 UTC

Official source (third-party copy, not published) archive/SG-08/2026-10-18/official.pdf
Retrieved 2026-10-18 06:48:39 UTC
SHA-256 2636e19ff1c86e862394d2fc900592e97b83c04cc35e3c8443108114b7f1dfba
Secondary source (third-party copy, not published) archive/SG-08/2026-10-18/secondary.html
Retrieved 2026-10-18 06:48:39 UTC
SHA-256 14f3b76f3e0e24dfb1fcaf1994b39eb7cb16def5ee161e99d860eb7726f5335f

Verification

B
Another official source, such as an announcement by the issuer, confirms the title, date and status. checked 2026-10-18 · alternative-official-url-fetched
Recheck of 18 Oct 2026. Stored and read: the framework's own text on imda.gov.sg, stored again in this run as the official source (same file as the record's earlier copy of 11 Oct 2026), whose cover reads 'Version 1.5 | Published 20 May 2026 (Updated 5 June 2026)' and which says the version incorporates feedback received since v1.0; and MDDI's press release of 22 Jan 2026 (mddi.gov.sg, copy of 18 Oct 2026; its text is the same as the copy of 11 Oct 2026 except the site footer), which gives the launch of Version 1.0 at Davos on 22 Jan 2026, IMDA as developer and the four dimensions. official_url now points to the framework's text and the press release becomes the secondary source. The framework's text confirms the title, the current version and its status, but not the date of Version 1.0, which rests on the press release, an announcement; so the date has another official source only (level B). No copy of Version 1.0 on an official host was found.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 8 field(s) supported.

History

  1. 2026-10-18verification.level: A → B — level computed from the evidence by tools/pipeline.py2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  2. 2026-10-18secondary_url: https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/arti → https://www.mddi.gov.sg/newsroom/singapore-launches-new-model-ai-governance-fram — the press release, previously official_url, becomes the secondary source2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  3. 2026-10-18official_url: https://www.mddi.gov.sg/newsroom/singapore-launches-new-model-ai-governance-fram → https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/arti — the record's official_url was MDDI's press release, an announcement; the framework's own text is published on imda.gov.sg and already stored as the record's secondary source, so it becomes official_url (rule 7c)2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  4. 2026-10-18notes: Minister Josephine Teo launched it at the WEF in Davos on 22 Jan 2026. The PDF c → Minister Josephine Teo announced Version 1.0 at the World Economic Forum in Davo — Version 1.5 and its changes from the stored PDF replace version 1's unconfirmed law-firm account of a May 2026 update; the 19 January 2026 publisher line (of a Version 1.0 PDF no longer stored) and the Committee of Supply study rest on no stored file and were removed2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  5. 2026-10-18summary: Voluntary governance framework for AI agents that act autonomously. It has four → Governance framework for organisations deploying AI agents, whether developed in — updated to Version 1.5: measures against automation bias, access controls and continuous monitoring added; 'voluntary' dropped, since no stored text uses the word2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  6. 2026-10-18official_identifier: Version 1.0 → – — 'Version 1.0' is a version label, which AGENTS.md section 5.6 excludes from official_identifier, and the stored PDF is Version 1.5; the framework has no citable number, so null2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  7. 2026-10-18date_last_updated: – → 2026-06-05 — the stored IMDA PDF is 'Version 1.5 | Published 20 May 2026 (Updated 5 June 2026)'; 5 June 2026 is the last change to the text, 20 May 2026 (Version 1.5) is in the notes2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)

Related records

Singapore Model AI Governance Framework Singapore · Published (non-binding, current)

Singapore MGF for Generative AI Singapore · Published (non-binding, current)

Citation

Infocomm Media Development Authority (IMDA). Model AI Governance Framework for Agentic AI. 2026-01-22. https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/artificial-intelligence/mgf-for-agentic-ai.pdf. Status: Published (non-binding, current); verified 2026-10-18 (level B). Local copy archive/SG-08/2026-10-18/official.pdf retrieved 2026-10-18 06:48:39 UTC, SHA-256 2636e19ff1c86e862394d2fc900592e97b83c04cc35e3c8443108114b7f1dfba. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record SG-08.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.