AI Regulation in Singapore: A 2026 Update

Article by Looi Teck Kheong, Global AI Ambassador and President of the Singapore Chapter at the Global Council for Responsible AI; Advocate & Solicitor of the Supreme Court of Singapore

Singapore continues to chart a distinctive path in AI governance, maintaining its pragmatic “soft law” approach that prioritizes industry innovation while establishing clear accountability frameworks. Unlike the European Union’s comprehensive AI Act, Singapore has not enacted dedicated, cross-sectoral AI legislation. Instead, the country relies on voluntary governance frameworks, sectoral guidelines, and targeted updates to existing laws to manage AI risks .

This strategy reflects Singapore’s broader ambition to position itself as a global hub for trusted AI development—balancing robust guardrails with the flexibility needed to attract investment and foster experimentation.

Major Development: The 2026 Model AI Governance Framework for Agentic AI

The most significant regulatory update in 2026 is the Model AI Governance Framework for Agentic AI, unveiled by the Infocomm Media Development Authority (IMDA) at the World Economic Forum in Davos on January 22, 2026 .

Why it matters: Agentic AI refers to systems capable of autonomous reasoning, planning, and action — AI that can independently execute tasks such as updating databases or initiating payments. While offering transformative efficiency gains, such systems introduce novel risks, including unauthorized actions, cascading errors, and diminished human oversight .

The Framework, a world-first in addressing this emerging technology, is structured around four core dimensions :

1. Assess and Bound Risks Upfront: Organizations must evaluate use cases based on factors such as the agent’s level of autonomy, access to sensitive data, and the reversibility of its actions. Design limits (e.g., restricting tool access) should be implemented early.
2. Ensure Meaningful Human Accountability: Humans must remain ultimately accountable. The Framework recommends establishing clear chains of responsibility and mandating human approval at significant checkpoints, particularly for high-stakes or irreversible decisions.
3. Implement Technical Controls: This includes safety testing, sandboxing, continuous monitoring, and deploying guardrails to prevent privilege escalation or misuse throughout the agent’s lifecycle.
4. Enable End-User Responsibility: Organizations must provide transparency on agent capabilities and limitations, along with user training to ensure effective oversight.
   

While voluntary, this Framework is expected to set a global benchmark, influencing both industry best practices and future regulatory thinking.

Supporting Developments: Strengthening Digital Infrastructure and AI Testing

Complementing the Agentic AI Framework, two additional developments reinforce Singapore’s governance ecosystem:

1. Proposed Digital Infrastructure Act (DIA)
   In February 2026, the government announced plans to introduce the Digital Infrastructure Act, expected to be tabled later in the year. The DIA will impose binding requirements on data centers and cloud service providers, including mandatory energy efficiency standards, cybersecurity measures, and incident reporting obligations. This marks a shift from advisory standards to legally enforceable rules for critical digital infrastructure.
2. Generative AI Evaluation Sandbox
   IMDA and the AI Verify Foundation launched the Generative AI Evaluation Sandbox in March 2026. This initiative brings together global model developers (Google, Anthropic, NVIDIA) and third-party testers to develop benchmarks for domain-specific and culturally appropriate AI evaluations. It aims to address gaps in current testing methodologies, particularly for Singapore’s multilingual context.
   

Implications for Organizations

For businesses deploying or developing AI in or from Singapore, the 2026 updates signal clear expectations:

• Accountability remains central: Even without a dedicated AI law, organizations remain liable under existing legislation such as the Personal Data Protection Act (PDPA) and sector-specific rules. The new frameworks provide guidance on how to mitigate that liability .
• Documentation and testing are key: Alignment with IMDA’s frameworks — particularly for agentic and generative AI — will serve as evidence of responsible practices.
• Prepare for evolving standards: The Agentic AI Framework is a “living document,” and IMDA actively invites feedback and case studies. Organizations should engage with consultations and monitor for future iterations

Conclusion

Singapore’s AI regulatory landscape in 2026 remains defined by adaptive governance: voluntary, principles-based frameworks that evolve with technology, supported by targeted infrastructure legislation. The new Model AI Governance Framework for Agentic AI represents a significant maturation of this approach, providing concrete guidance for one of the most complex frontiers in artificial intelligence. For now, the emphasis is on responsible innovation—trust as a competitive advantage, not a compliance burden.

Here are the references that support the regulatory briefing on AI legislation in Singapore, organized by the key developments cited.