SG unveils model governance framework for agentic AI at WEF

Singapore has introduced a model AI governance framework for agentic AI, developed by the Infocomm Media Development Authority (IMDA), at the World Economic Forum on January 22, 2026.

The framework is the first to offer comprehensive guidance for organizations on deploying agentic AI responsibly, emphasizing human accountability and risk mitigation.

Unlike traditional AI, agentic AI can reason and perform multi-step tasks, raising concerns over data security and autonomous actions such as database updates or payments.

The framework outlines measures across four areas: risk assessment, human oversight, technical controls, and transparency.

It was created with input from government and private sector stakeholders and is intended to be a living document, welcoming feedback and case studies.

Singapore aims to foster a trusted AI ecosystem through initiatives like this framework and regional collaborations within ASEAN.

The IMDA continues to develop guidelines to support safe and reliable AI deployment.

🔗 Source: Infocomm Media Development Authority

🧠 Food for thought

Implications, context, and why it matters.

Unclear how quickly organisations are expected to implement the framework

The framework is described as guidance for organisations deploying agentic AI, but it is not stated whether compliance will be enforced or required.
It remains undefined whether the framework will be required for Singapore government contracts or treated as a minimum bar in regulated industries.
IMDA calls the framework a “living document” and invites feedback and case studies, but no schedule is given for adoption or updates.
Because timing and scope are left open, urgency and staffing needs are harder to estimate for the technical and non-technical steps being recommended.

Signal for AI safety vendors, though the announcement does not set a “standard”

AI safety and assurance vendors can treat the framework as a demand signal for responsible agentic AI rollouts, even though it is not described as a binding standard.
IMDA is developing testing guidelines for agentic AI applications. This work builds on its “Starter Kit Testing of LLM-Based Applications for Safety and Reliability”. It also lists the AI Verify toolkit (a software tool for testing and validating AI systems) as part of Singapore’s trustworthy AI ecosystem initiatives. 1
Vendors can use these resources to build product features that help enterprises apply the framework’s controls, including risk boundaries plus meaningful human accountability and oversight.
Singapore is leading the ASEAN Working Group on AI Governance (a regional group focused on AI governance across Southeast Asia) to develop a trusted AI ecosystem within ASEAN, which gives regional weight to solutions aligned with the framework. 1