Singapore unveiled the world's first comprehensive governance framework for agentic AI systems, targeting AI agents capable of autonomous reasoning, planning, and action. The Model AI Governance Framework for Agentic AI was announced by Singapore's Minister for Digital Development and Information, Josephine Teo, at the World Economic Forum in Davos, and it provides organisations with structured guidance on deploying AI agents responsibly while maintaining meaningful human oversight.
Developed by the Infocomm Media Development Authority (IMDA), the framework builds upon the original Model AI Governance Framework introduced in 2020, but specifically addresses the distinct risks of agentic AI, including unauthorised actions, data leakage, and automation bias. The framework is structured around four governance dimensions: assessing and bounding risks upfront, ensuring meaningful human accountability through defined approval checkpoints, implementing technical controls such as sandboxing and continuous monitoring throughout the agent lifecycle, and promoting end-user responsibility via transparency and training. April Chin, Co-CEO of AI assurance firm Resaro, stated that the framework fills a critical gap in policy guidance by helping organisations define agent boundaries, identify risks, and implement mitigations such as agentic guardrails. Although the framework is voluntary and nonbinding, it is expected to shape global norms, particularly alongside complementary tools such as the AI Verify toolkit and the Association of Southeast Asian Nations governance initiatives. Earlier, in October 2025, the Cyber Security Agency of Singapore released a dedicated addendum to its AI security guidelines, specifically addressing the unique risks posed by agentic AI systems.
With the launch of the Model AI Governance Framework for Agentic AI, Singapore has reinforced its pioneering role in AI regulation, building on the foundations laid by the 2020 framework and subsequent initiatives. The IMDA has designated the document as a living framework, actively seeking industry feedback and case studies to refine its guidelines, while also developing dedicated testing standards for agentic AI applications.
Sources:
1.
Aaron Tan,
2.
3.
Lucas Nicolet-Serra
