# AI rules in Singapore

Singapore relies on voluntary model governance frameworks and testing tools (including the first agentic-AI framework) plus targeted laws on election deepfakes and online harms, with MAS AI risk guidelines for finance pending.

- **Approach:** Voluntary principles
- **Key bodies:** MDDI, IMDA, PDPC, AI Verify Foundation, MAS, Online Safety Commission
- **Strategy:** [National AI Strategy 2.0](https://www.mddi.gov.sg/newsroom/04122023/) (2023)
- **Last reviewed:** 4 Oct 2026

## Rules (5)
- [MGF for Agentic AI](https://wheresthe.ai/rule/sg-imda-model-ai-governance-framework-agentic-ai-2026/index.md): In force, Guideline, effective 22 Jan 2026. Voluntary framework, launched at the World Economic Forum in Davos, guiding organisations on deploying AI agents responsibly, with technical and non-technical measures to manage risks such as unauthorised or erroneous actions, while stressing that humans remain accountable.
- [MGF for GenAI](https://wheresthe.ai/rule/sg-model-ai-governance-framework-generative-ai-2024/index.md): In force, Guideline, effective 30 May 2024. Voluntary framework setting out nine dimensions for trustworthy generative AI (e.g. accountability, data, trusted development, incident reporting, testing and assurance, security, content provenance, safety R&D, AI for public good). It extends Singapore's 2019/2020 Model AI Governance Framework for traditional AI.
- [MAS AI Risk Management Guidelines (AIRG)](https://wheresthe.ai/rule/sg-mas-ai-risk-management-guidelines-consultation-2025/index.md): In consultation, Guideline. Proposed supervisory guidelines requiring financial institutions to identify and inventory AI use, assess risk materiality, and apply proportionate controls across the AI lifecycle, with board and senior-management oversight. MAS proposed a 12-month transition after issuance.
- [Election deepfake ban](https://wheresthe.ai/rule/sg-elections-online-advertising-deepfake-amendment-2024/index.md): In force, Law. Prohibits publishing, boosting, sharing or reposting online election advertising containing realistic digitally generated or manipulated content (including AI deepfakes) that shows a candidate saying or doing something they did not. The ban runs from the issue of the writ of election until close of polling; returning officers can issue corrective directions.
- [OSRA Act / Online Safety Commission](https://wheresthe.ai/rule/sg-online-safety-relief-accountability-act-2025/index.md): Partly in force, Law, effective 29 Jun 2026. Creates a Commissioner of Online Safety and Online Safety Commission that can order platforms and users to act on 13 categories of online harm, and new statutory torts for victims. The first phase (from 29 Jun 2026) focuses on five harms including intimate image abuse and image-based child abuse, which cover AI-generated (deepfake) imagery.

## Programmes and facts
- Model AI Governance Framework for Agentic AI launched 22 Jan 2026
- AI Verify Foundation runs the AI Verify testing framework and Global AI Assurance Sandbox
- Online Safety Commission began operations on 29 Jun 2026

---
Canonical page: https://wheresthe.ai/j/sg/ · Not legal advice.