Singapore has formally introduced a new governance framework aimed at guiding the safe and accountable use of agentic artificial intelligence systems as these technologies gain traction across industries. The framework responds to rising concerns around security operational control and accountability linked to AI systems that are capable of independent reasoning planning and action without direct human involvement.
The Model AI Governance Framework for Agentic AI was announced on 22 January 2026 by Singapore’s minister for digital development and information Josephine Teo during the World Economic Forum held in Davos Switzerland. Developed by Infocomm Media Development Authority the framework builds on Singapore’s earlier AI governance guidelines introduced in 2019 but places specific emphasis on agentic AI systems that go beyond content generation. Unlike generative AI tools these systems are designed to plan across multiple steps interact with digital environments and execute tasks such as updating customer records or processing financial transactions autonomously. IMDA highlighted that while these capabilities can significantly enhance business productivity they also introduce new risks related to unauthorised actions unintended outcomes and reduced clarity over human accountability.
IMDA noted that the increased autonomy of agentic AI systems can amplify challenges such as automation bias where users may over trust systems that have previously performed reliably. This dynamic raises concerns about oversight decision validation and responsibility when errors occur. The framework is intended to guide organisations deploying agentic AI whether developed internally or sourced from third party providers. It outlines governance measures to support safe and effective deployment including defining clear limits on an agent’s authority access to data and use of tools. It also encourages the establishment of checkpoints that require human approval at critical stages to mitigate the risk of unchecked automation. Baseline testing continuous monitoring across the AI lifecycle and transparency around when users are interacting with AI agents are also key elements of the guidance. IMDA emphasised the importance of training personnel to supervise agentic systems to ensure informed human oversight remains central to operations.
The development of the framework involved consultation with major technology firms and assurance providers reflecting broad industry engagement. Elsie Tan country manager for worldwide public sector at Amazon Web Services stated that as agentic AI systems begin making decisions with real world consequences organisations need concrete mechanisms for visibility containment and alignment embedded into infrastructure alongside human judgement. April Chin co CEO of AI assurance firm Resaro described the framework as addressing a critical policy gap by helping organisations define agent boundaries identify risks and implement safeguards such as agentic guardrails. Serene Sia country director for Singapore and Malaysia at Google Cloud highlighted that trust in agentic AI is a shared responsibility and noted Google’s work on open standards such as Agent2Agent Protocol and Agent Payments Protocol to support secure interoperable multi agent systems. IMDA has characterised the framework as a living document and is inviting industry feedback and case studies to refine its guidance. The authority is also preparing more detailed testing guidelines for agentic AI applications building on its existing starter kit for testing large language model based systems. This follows an earlier move in October 2025 when Cyber Security Agency of Singapore released an addendum focused on securing agentic AI systems outlining practical controls and methods for assessing risks by mapping workflows where vulnerabilities could be exploited.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
