Singapore Debuts Global First Governance Framework for Autonomous Agentic AI

During the World Economic Forum in Davos on January 22, 2026, Minister for Digital Development and Information Josephine Teo announced the launch of the Model AI Governance Framework for Agentic AI. Developed by the Infocomm Media Development Authority, this framework is the first in the world to address the unique challenges of AI agents that can reason and execute tasks independently. It marks a significant shift from previous guidelines that focused primarily on content generation, moving instead toward the management of autonomous actions. Unlike standard generative models, AI agents possess the capability to update databases and process payments without direct human prompts. The new framework provides a structured overview of technical and non-technical measures to mitigate risks such as unauthorized transactions or automation bias. By establishing clear guardrails, the Singapore government aims to foster a trusted environment where businesses can safely harness these advanced capabilities for digital transformation.

Bounding Risks and Establishing Control

The framework outlines four critical dimensions for responsible deployment, starting with the assessment and bounding of risks. Organizations are advised to "box in" an agent's power by limiting its access to sensitive tools and defining strict operational parameters. This proactive approach ensures that AI agents operate within a controlled scope, preventing them from performing high-stakes tasks without the necessary maturity or testing. Maintaining meaningful human accountability remains a core pillar of the new guidance. The framework mandates significant checkpoints where human approval is required, particularly for actions that could have financial or legal consequences. By defining these boundaries, the IMDA seeks to counter the psychological tendency for humans to over-trust automated systems that have worked reliably in the past.

Global Leadership in AI Safety

Technical controls and lifecycle management form another essential part of the guidelines. Companies are encouraged to implement rigorous baseline testing and restricted access to whitelisted services throughout the agent's life. These measures are designed to detect unintended behaviors or "drift" after a system goes live, ensuring that the agents remain reliable and safe for long-term enterprise use. Singapore is actively working with international partners through its AI Safety Institute and the ASEAN Working Group on AI Governance to harmonize these standards. This latest initiative follows the successful rollout of the 2020 Model Governance Framework and the 2024 Generative AI guidelines. The goal is to build a global ecosystem where autonomous AI is recognized as a trusted and secure tool for global productivity.