Singapore's Model AI Governance Framework is guidance, not law: the Infocomm Media Development Authority (IMDA) and the Personal Data Protection Commission (PDPC) publish it to help private-sector organisations deploy AI responsibly, while the binding duties that reach AI marketing come from legislation such as the Personal Data Protection Act 2012 (PDPA). The guidance comprises three related frameworks: traditional AI, generative AI and agentic AI. The Model Framework for traditional AI dates from 23 January 2019, with a second edition on 21 January 2020. The Model AI Governance Framework for Generative AI, from IMDA and the AI Verify Foundation, followed on 30 May 2024 with nine dimensions, content provenance among them. The Model AI Governance Framework for Agentic AI, launched on 22 January 2026, adds four dimensions for AI agents and states that humans are ultimately accountable. AI Verify, IMDA's testing framework and toolkit, sits alongside them.
What the Model AI Governance Framework Covers, and Who It Applies To
The Model AI Governance Framework is written for private-sector organisations that deploy AI, and it gives them detailed, readily implementable guidance on the ethical and governance issues AI raises. The PDPC released the first edition on 23 January 2019 "for broader consultation, adoption and feedback", and the PDPC's AI governance page states that the framework aims to promote public understanding and trust by explaining how AI systems work, building good data accountability practices and creating open and transparent communication. The second edition, released on 21 January 2020, added considerations such as robustness and reproducibility and expanded the section on interactions with stakeholders. It takes a sector-agnostic and technology-agnostic approach that complements sector-specific requirements and guidelines.
Is the Model AI Governance Framework legally binding?
No: the Model Framework is voluntary guidance, and the PDPC encourages organisations to use it, with its companion guides, for internal discussion and implementation. The PDPC calls the framework a living document that evolves in future editions, and it invites trade associations, chambers and professional bodies to adapt it for their own use. IMDA describes its companion Starter Kit for testing LLM-based applications, and its Transparency Guidelines for Generative AI Chatbots, as voluntary guidelines. The legal duties around AI and personal data come from the PDPA, and the PDPC's advisory guidelines explain how it applies to AI systems; sector regulators add their own expectations on top.
What are the framework's guiding principles and its four areas?
The Model Framework rests on two guiding principles: decisions made by AI are explainable, transparent and fair, and AI systems are human-centric. It turns those principles into practice in four areas:
- Internal governance structures and measures: clear roles and responsibilities, standard operating procedures to monitor and manage risks, and staff training.
- The level of human involvement in AI-augmented decision-making: an appropriate degree of human involvement that keeps the risk of harm to individuals low.
- Operations management: minimising bias in data and models, with a risk-based approach to explainability, robustness and regular tuning.
- Stakeholder interaction and communication: making AI policies known to users, letting users give feedback where possible, and keeping communications easy to understand.
What does the generative AI framework add?
The Model AI Governance Framework for Generative AI sets out nine dimensions for a trusted generative AI ecosystem and offers practical suggestions as first steps, building on the framework for traditional AI. The AI Verify Foundation and IMDA released a proposed version for international views on 16 January 2024 and the final framework on 30 May 2024. The nine dimensions are accountability, data, trusted development and deployment, incident reporting, testing and assurance, security, content provenance, safety and alignment research and development, and AI for public good. The Foundation describes content provenance as transparency about where content comes from, as a useful signal for end-users.
What does the agentic AI framework add?
The Model AI Governance Framework for Agentic AI, launched by Singapore's Minister for Digital Development and Information at the World Economic Forum on 22 January 2026, guides organisations that deploy AI agents and states that humans are ultimately accountable for them. IMDA's launch release sets out four dimensions: assessing and bounding the risks upfront, including limits on an agent's autonomy and its access to tools and data; making humans meaningfully accountable through checkpoints at which human approval is required; implementing technical controls and processes across the agent life cycle; and enabling end-user responsibility through transparency and training. IMDA updated the framework in May 2026 with real-world case studies and new practices on multi-agent systems, third-party agents and automation bias.
Where do AI Verify and ISAGO fit?
AI Verify is IMDA's AI governance testing framework and software toolkit, and ISAGO is the companion guide that helps organisations assess how well their practices align with the Model Framework. IMDA states that the AI Verify framework sets out 11 governance principles, aligns with international AI standards, was released for an international pilot in May 2022, was open-sourced in 2023 and now supports testing for both traditional and generative AI use cases. The AI Verify Foundation, which IMDA established as a not-for-profit body, has nine premier members and more than 180 general members. ISAGO, the Implementation and Self-Assessment Guide for Organisations, was developed with the World Economic Forum's Centre for the Fourth Industrial Revolution and contributions from more than 60 organisations.
General guidance, not legal advice.
What the Framework Means for Your Marketing and SEO Content
For a marketing team, the framework's most direct points are transparency and accountability: tell people how your business uses AI, keep a named person accountable for what AI produces, and be open about where content comes from. These are governance recommendations. PDPA duties and sector requirements apply according to the personal data processed, the activity and the organisation. The table maps common AI marketing uses to the guidance behind each one.
| Claim type | What the rule says | What to check | Source |
|---|---|---|---|
| AI-assisted articles, images and video | The generative AI framework names content provenance, transparency about where content comes from, as one of nine dimensions | Whether your content policy records or discloses AI assistance | AI Verify Foundation and IMDA, MGF for Generative AI |
| Statements about your AI use | The Model Framework's stakeholder communication area: make AI policies known to users and keep communications easy to understand | An AI policy page written in plain language | PDPC, Model AI Governance Framework |
| Customer feedback on AI features | Users have a way to give feedback, where possible | A visible feedback or contact route beside AI features | PDPC, Model AI Governance Framework |
| A customer-facing chatbot | IMDA's voluntary transparency guidelines describe a chatbot info card covering what the chatbot does and does not do, how reliable and safe it is, how user data is handled and how to report issues, published at launch and kept up to date | An info card on the chatbot's page | IMDA, Transparency Guidelines for Generative AI Chatbots |
| Accountability for published AI content | Clear roles and responsibilities inside the organisation; the generative AI framework's accountability dimension aims at players being responsible to end-users | A named approver for every AI-assisted page | PDPC Model Framework; MGF for Generative AI |
| Accuracy of AI-drafted claims | IMDA's Starter Kit for testing LLM-based applications covers five common risks, hallucination and inaccuracy first among them | Every AI-drafted fact checked against an approved source before publication | IMDA, Starter Kit for Testing LLM-based Applications |
| AI agents that publish or send content | The agentic framework places limits on agents' powers and sets checkpoints at which human approval is required | Human approval before an agent publishes, emails or posts | IMDA, MGF for Agentic AI |
| Personal data in AI personalisation | The PDPA applies to the collection and use of personal data in AI systems; deployers of generative AI systems bear primary responsibility for PDPA compliance | Consent, notification and purpose for each data use | PDPC advisory guidelines on AI |
Sources: PDPC, Singapore's Approach to AI Governance; AI Verify Foundation, Model AI Governance Framework for Generative AI; IMDA, Artificial Intelligence in Singapore; IMDA, Singapore Launches New Model AI Governance Framework for Agentic AI, 22 January 2026; PDPC, guidance on the use of personal data in generative AI. All checked 24 October 2026.
Does the framework require labels on AI-generated content?
No: the generative AI framework names content provenance as one of its nine dimensions and offers practical suggestions, not a labelling rule. How your business shows where content comes from, whether through a disclosure line, an AI policy page or internal records, stays a decision for your organisation. Sector rules still apply to the content itself: a financial product advertisement drafted with AI remains subject to applicable MAS advertising requirements, as the MAS financial advertising guide explains.
How does the PDPA apply to AI-driven marketing?
The PDPA is law, and the PDPC's advisory guidelines explain when organisations use personal data to develop and deploy AI systems, from recommendation engines to generative AI. The PDPC's guidance on generative AI states that where data users provided for other purposes is used to develop a generative AI model and no exception to consent applies, organisations obtain consent through AI-specific notifications that explain the purpose, the data used and how individuals decline or withdraw consent. It also states that system deployers bear primary responsibility for PDPA compliance, including clear purposes for processing and regular reviews of safeguards, especially for agentic AI systems. The PDPA and DNC marketing guide covers consent for marketing messages.
General guidance, not legal advice.
What the Framework Means for AI Answers About Your Business
The framework is written for organisations that develop and deploy AI, so it shapes how your business uses AI rather than what ChatGPT, Claude, Gemini or Perplexity say about your business. In Aemorph's working model, those AI assistants describe a business from the public sources they find, your own pages and profiles among them, so the accuracy of their answers starts with the facts you publish. IMDA's advisory for individuals on generative AI makes the same point from the reader's side: verify content received from generative AI before using or sharing it.
What does "humans are ultimately accountable" mean for AI-assisted content?
In Aemorph's recommended workflow, the agentic framework's accountability principle becomes a named person who approves what AI drafts before it reaches customers. The framework itself defines significant checkpoints at which human approval is required, rather than approval of every output, and IMDA's launch release warns of automation bias, the tendency to over-trust an automated system that has performed reliably in the past. A content workflow that applies the principle names the approver for each page, records the source behind each fact and keeps that record when the page changes.
How does the framework connect to sector rules?
Sector regulators build on the national framework rather than replace it: MAS's consultation on its Guidelines on Artificial Intelligence Risk Management described them as complementing IMDA's Model AI Governance Framework and MAS's own FEAT principles. MAS issued those guidelines on 7 October 2026, they take effect on 7 October 2027, and they apply to all financial institutions and all forms of AI. IMDA's AI Verify toolkit also hosts sector plugins, including MAS's Veritas Toolkit for assessing AI models against the FEAT principles and the Competition and Consumer Commission of Singapore's AI Markets toolkit.
General guidance, not legal advice.
How Aemorph Checks AI-Assisted Content Before Publication
Aemorph traces each fact it publishes to an approved source and a named approver, and your nominated point of contact signs off every regulated claim before Aemorph publishes it. Aemorph is not a law firm. It does not certify that your organisation aligns with the Model AI Governance Framework, it does not run AI Verify tests on your systems, and it does not confirm that your content meets the PDPA or any sector rule: how your organisation applies the framework, and its legal obligations, stay with your organisation.
| Check | Applies to | Standard | Checked against | When |
|---|---|---|---|---|
| AI use and AI-generated content | Every client | Each published fact traces to an approved source and a named approver | Your point of contact's sign-off | Before Aemorph publishes |
| Medical and health claims | Hospitals and healthcare providers | Conservative wording, no exaggerated promises | PubMed, professional associations, government health authorities | Before Aemorph publishes |
| Financial and insurance claims | Financial services and insurance clients | Disclaimers present, no promised financial outcome | Your point of contact's sign-off | Before Aemorph publishes |
| Client data and access | Every client | Client-owned assets and named least-privilege accounts | ISO 27001-aligned practice | From kick-off |
Source: Aemorph compliance practice, October 2026.
Who signs off the content on your pages?
You decide who signs off: the business owner, the chief marketing officer or the risk team. Aemorph publishes a regulated claim only after that person approves it, and the facts in your verified record are the facts your business supplies and approves. Inside the Showstopper Sensor (Sprint 1), a 30-day showstopper-answer sprint for ChatGPT, Claude, Gemini and Perplexity, the same sign-off covers the showstopper answers before they go live. Aemorph's privacy policy names the Personal Data Protection Act 2012.
General guidance, not legal advice.
Singapore Businesses Building AI Visibility: Start With a Free Strategic Conversation
Singapore businesses that want accurate answers about them in AI chats, built on facts their own approvers have signed off, start with a free strategic conversation with Aemorph. Aemorph is headquartered in Singapore and replies to every message within 24 hours. The LLM Influencer System page sets out how Aemorph runs SEO, AEO and GEO as one programme on one verified record, and the free Hallucination Report compares what the AI engines say about your business with what your website currently says.