TLDR: Singapore’s AI Verify Foundation and the Infocomm Media Development Authority (IMDA) have launched a global AI-assurance sandbox. This initiative expands on a successful pilot, enabling more companies and regulators worldwide to rigorously test real-world AI applications. The sandbox aims to address emerging AI risks, inform policy development, and establish accreditation for AI testers, reinforcing Singapore’s commitment to responsible AI innovation.
Singapore has taken a significant step forward in global AI governance and assurance with the announcement of a new global AI-assurance sandbox by the AI Verify Foundation (AIVF) and the Infocomm Media Development Authority (IMDA) on July 7, 2025. This initiative is designed to provide a robust platform for a wider array of companies and regulators to test Artificial Intelligence (AI) applications in real-world scenarios.
Building upon the insights garnered from a successful pilot phase launched earlier this year at the Paris AI Action Summit in February 2025, the expanded sandbox incorporates new AI archetypes, including agentic AI. It also specifically targets emerging risks such as data leaks and vulnerabilities to prompt injections, reflecting the evolving landscape of AI development and deployment. The pilot, which ran between March and May 2025, saw 17 organizations across 10 diverse sectors, including human resources, healthcare, and finance, have their Generative AI (GenAI) applications assessed by specialist testing firms. Senior Minister of State for Digital Development and Information, Tan Kiat How, highlighted on May 29 that this pilot demonstrated how GenAI applications perform under practical conditions, with the aim of making GenAI use ‘boring and predictable’ to ensure reliability.
A key aspect of this global sandbox is its openness to sector regulators. This allows them to develop and receive real-life feedback on their AI governance and testing guidelines, fostering a collaborative environment for establishing best practices. The insights gathered from the sandbox will be instrumental in informing future policy guidance and are intended to pave the way for the development of accreditation for AI testers, thereby standardizing and professionalizing the field of AI assurance.
To support these efforts, the sandbox will leverage IMDA’s recently launched Starter Kit for Safety Testing of Large Language Models (LLMs) based applications. This integration aims to provide a more standardized and structured operational framework for testers, aligning with Singapore’s pragmatic, risk-based approach to AI. This approach seeks to maximize innovation and adoption while implementing appropriate guardrails to cultivate a trusted AI ecosystem.
Also Read:
- Proactive Vulnerability Testing: The Rise of AI Red Teaming for Enhanced Safety
- Geneva to Host 7th AI for Good Global Summit, Driving AI for Sustainable Development
The AI Verify Foundation, established in June 2023 as a not-for-profit subsidiary of IMDA, plays a pivotal role in this endeavor. Its mission is to boost AI testing capabilities and assurance globally by fostering a community that contributes to AI testing frameworks, code bases, standards, and best practices. The Foundation also serves as a neutral platform for open collaboration and idea-sharing on AI testing and governance, and it actively nurtures a network of advocates for AI, driving broad adoption of AI testing through education and outreach. The Foundation boasts nine premier members, including industry giants like AWS, Dell, Google, IBM, Microsoft, Red Hat, Resaro, and Salesforce, alongside over 180 general members, underscoring its broad industry support and collaborative nature. The initiative also emphasizes the critical role of human experts at every stage of testing, from designing appropriate tests to interpreting results, acknowledging that while AI technology advances, a human touch remains essential for ensuring trustworthiness.
