Singapore has introduced a new Model AI Governance Framework for agentic AI, providing a structured approach to responsible AI deployment. This first-of-its-kind framework, announced at the World Economic Forum, offers guidance on mitigating risks while fostering innovation, emphasizing that human accountability remains paramount in an increasingly automated world.
As organizations globally race to harness the power of artificial intelligence, the rise of agentic AI presents both immense opportunities and significant challenges. Unlike traditional AI, agentic systems can independently reason and act on a user's behalf, automating complex workflows from customer service to enterprise resource planning. However, this autonomy introduces new risks, including the potential for unauthorized actions and the erosion of human oversight. Singapore’s new AI governance framework directly addresses this critical challenge, offering a blueprint for organizations, companies, and multinational organizations to innovate responsibly—supporting ethical use and a balanced approach that can safeguard consumer interests across industries adopting data-driven technologies.
A Closer Look at the Agentic AI Governance Framework
Developed by the Infocomm Media Development Authority (IMDA), the framework builds upon Singapore's established leadership in AI governance and the country’s evolving AI governance approach. It provides a clear, four-dimensional approach to help organizations navigate the complexities of agentic AI, ensuring that progress does not come at the expense of safety and trust. This aligns with the nation's broader Singapore's AI regulation approach, which balances innovation with robust guardrails—including practical governance integration that helps translate AI governance requirements into core organizational practice.
In practice, the framework can also complement related national ecosystem efforts such as the AI Verify Foundation, and can be read alongside guidance and expectations from other Singapore stakeholders (e.g., PDPC and MAS) depending on a sector-specific approach and risk profile, rather than relying solely on one-time assessments.
| 1. Risk Assessment | Organization are guided to assess and bound risks upfront by selecting appropriate use cases and replacing clear limits on an agent's autonomy, access to tools, and data. |
| 2. Human Accountability | The framework emphasizes defining significant checkpoints where human approval is required, ensuring meaningful human oversight and accountability for an agent's actions. |
| 3. Technical Controls | It recommends implementing technical controls throughout the agent's lifecycle, including rigorous baseline testing and controlled access to whitelisted services to prevent misuse. |
| 4. End-User Responsibility | Promoting transparency and providing education and training for end-users are highlighted as crucial steps to ensure responsible interaction with agentic systems. |
This structured approach provides a practical pathway for organizations to implement effective AI management systems and broader AI governance infrastructures that can adapt to the evolving capabilities of agentic AI. For some private sector organisations, it may also pair well with an AI governance testing framework (and separate implementation approaches by function or business unit) to achieve comprehensive visibility into agent behavior in the digital domain.
Balancing Innovation and Accountability in AI Governance
The launch of this framework at the World Economic Forum signals a global recognition of the need for proactive AI governance. As April Chin, Co-Chief Executive Officer of Resaro, noted, the framework “fills a critical gap in policy guidance for agentic AI” by establishing foundational principles for assurance and risk mitigation. It provides a model for how to build trust in AI systems by embedding accountability into their design and deployment from the outset—supporting AI innovation without defaulting to mandatory regulatory frameworks.
This initiative is part of a broader global conversation around AI regulation, with many jurisdictions developing their own approaches. For example, the framework complements other significant regulatory efforts like the EU AI Act, which also seeks to establish a comprehensive legal framework for trustworthy AI. By sharing its model framework, Singapore contributes to the development of harmonized international AI governance standards and best practices, helping businesses and regulators interpret international rules consistently—even as AI governance regulations evolve at different speeds globally.
The Path Forward for Responsible AI Deployment
For businesses and technology leaders, Singapore's new AI governance framework offers more than just a set of rules; it provides a competitive advantage. By adopting a structured approach to AI risk management, organizations can build greater trust with customers, partners, and regulators. This proactive stance on governance is becoming increasingly important as AI systems become more integrated into critical business operations, and it can offer governance insights and strategic insights that inform long-term strategic directions.
As the AI landscape continues to evolve, with similar AI governance frameworks in the Asia-Pacific region and beyond, the principles of transparency, accountability, and human oversight will remain central to sustainable innovation. This new framework from Singapore provides a valuable and timely resource for any organization committed to deploying AI in a manner that is both powerful and principled—whether through a voluntary compliance model, participation in emerging assurance initiatives such as a global AI assurance pilot, or guidance shaped by an advisory council and other stakeholders.
