Singapore unveils 'Agentic AI' governance framework
Singapore has released a new governance framework specifically for "Agentic AI," systems capable of autonomous decision-making. The framework is described as a first-of-its-kind initiative, positioning Singapore as a regional leader in regulating next-generation AI. The government engaged with industry, academia, and civil society, creating a model that could influence digital policy across the ASEAN region.
- The framework, officially titled the "Model AI Governance Framework for Agentic AI," was unveiled by Singapore's Infocomm Media Development Authority (IMDA) at the World Economic Forum in Davos on January 22, 2026. This initiative builds upon Singapore's original Model AI Governance Framework, which was first released in 2019 and updated in 2020. - The governance model is structured around four key pillars: assessing and bounding risks upfront, ensuring meaningful human accountability, implementing technical controls, and promoting end-user responsibility. It is designed as a "living document" to evolve with technology and stakeholder feedback. - Unlike generative AI, agentic AI can take autonomous actions, interact with other systems, and execute multi-step tasks with limited human intervention, introducing unique risks such as unauthorized actions, cascading failures, and challenges in tracing accountability. The framework specifically addresses these higher-stakes risks. - While the framework is voluntary and not legally binding, it aims to establish global norms and best practices for organizations developing or deploying agentic AI. It complements other national initiatives like the AI Verify toolkit and the National AI Strategy 2.0, which has allocated over SGD 1 billion over five years to advance AI capabilities. - The IMDA developed the framework in consultation with industry partners, including tech companies like Google Cloud and Amazon Web Services, and AI assurance firms such as Resaro. - The framework recommends practical measures such as using sandboxed environments for testing, establishing clear checkpoints for human approval on high-stakes actions, restricting agent access to sensitive data and external tools, and ensuring end-users are aware they are interacting with an AI agent. - This initiative positions Singapore alongside other regional efforts in AI governance, such as South Korea's AI Basic Act (2026) and Taiwan's AI Basic Act (2025), contributing to a convergence of regulatory norms in the Asia-Pacific region.
