Singapore: Ministry of Digital Development and Information announced Model AI Governance Framework for Agentic AI (Version 1.0)

On 22 January 2026, the Ministry of Digital Development and Information (MDDI) released the Model AI Governance Framework for Agentic AI (Version 1.0). The Model AI Governance Framework for Agentic AI was developed by the Infocomm Media Development Authority (IMDA). The framework applies to organisations that develop or deploy agentic AI systems capable of multi-step planning, autonomous action-taking, adaptation to new information, and interaction with other agents and external systems. The framework builds on the Model AI Governance Framework and sets out governance considerations for agentic AI across the design, development, deployment, and post-deployment lifecycle. It identifies four areas of focus. These are assessing and bounding risks upfront, making humans meaningfully accountable, implementing technical controls and processes, and enabling end-user responsibility. The framework specifies agent-related risk sources and risk types, including erroneous actions, unauthorised actions, biased or unfair outcomes, data breaches, and disruption to connected systems, including cascading effects in multi-agent setups. It sets expectations for defining agent limits on tools, data access, autonomy, and scope of impact. It establishes agent identity and permissions and allocates organisational responsibilities across the agent lifecycle. The guidance also defines human oversight checkpoints for high-risk or irreversible actions, requires pre-deployment testing, and calls for gradual deployment. It mandates continuous monitoring, logging, and auditing, and ensures transparency, information, and training for users interacting with or supervising agentic AI systems.