Resource Library

Released by IMDA, this is a practical reference for enterprises and testers, which brings together emerging good practices for testing agentic systems for safety and reliability. It builds on IMDA's Starter Kit for LLM Apps, and focuses on the additional testing needs introduced by agents. It draws on learnings from the Global AI Assurance Sandbox, Singapore AISI's testing work, and consultations with government agencies and over 20 industry players.

This is intended to be a living document, and can be updated as testing practices mature and new insights emerge. We welcome feedback and case studies. Share your feedback with us to let us know what works and what can be improved.

This framework provides guidance to organisations on how to deploy AI agents responsibly, while emphasising that humans are ultimately accountable. It gives organisations a structured overview of the risks of agentic AI and emerging best practices in managing these risks. The framework was updated in May 2026 to include real-world case studies on how organisations can operationalise the framework’s recommendations to meet their needs.

Released by IMDA, this Starter Kit consolidates emerging best practices from industry and government into practical, voluntary guidelines for testing GenAI apps – applicable across sectors and use-cases. It helps businesses by providing a step-by-step guide on how to identify risks, apply robust testing methodologies, and assess whether their applications meet baseline safety and reliability. It covers five common risks: hallucination and inaccuracy, bias in decision-making, undesirable content, data leakage, and vulnerability to adversarial prompts.

Singapore and Rwanda have introduced the world’s first AI Playbook for Small States to shape inclusive global discourse on harnessing the potential of AI. First announced at the Asia Tech x Singapore Summit 2024 on 30 May 2024, the AI Playbook is developed by Singapore’s Infocomm Media Development Authority (IMDA) in collaboration with Rwanda’s Ministry of Information Communication Technology (ICT) and Innovation, with consultations taking place with Digital Forum of Small States (“Digital FOSS”) members since the start of the year.

The Playbook addresses common challenges small states face in adopting and harnessing the potential of AI, such as availability of resources and funding, access to data and AI talent, fostering a trusted environment such as developing holistic frameworks and practical testing tools such as AI Verify.

Given the rapidly evolving nature of AI, the Playbook is shaped as a living document that continuously pulls together the collective experiences and strategies from small states.

More information and the AI Playbook itself can be found in the following links.

Understanding and implementing responsible AI is a continuous journey, and every organisation is at a different stage. Whether you’re just beginning to explore AI governance or you’re looking to refine an established framework, our resources are designed to guide you through every step of this important process.

Singapore’s collection of guidance documents and tools offers practical insights and actionable practices tailored to your organisation’s level of maturity in AI governance. From foundational frameworks for those starting out, to advanced practices and technical tests for more sophisticated AI implementations, our materials provide the support you need to develop and enhance your AI governance capabilities.

Use our flowchart to assess where your organisation currently stands in its responsible AI journey. This will help you identify the right resources and tools to move forward confidently and responsibly.

Looking to test your AI models? Explore how AI Verify and Project Moonshot can help you assess and enhance your AI systems.

Find out which evaluation tools are best suited to your needs and how they can help you align with key AI governance principles and build trust with your stakeholders.

Uncover the principles of responsible AI and learn how to implement them with AI Verify. In this video, we provide a step-by-step walkthrough of AI Verify toolkit, designed to help you master its features and functionalities. Whether you are a beginner or looking to test your AI models, this workshop covers:

1. Introduction to AI Verify
2. Basic Operations and Features

By the end of video, you will have the knowledge to use AI Verify toolkit. If you are a developer looking to contribute to AI Verify, please watch Part II.

Learn to create custom plugins to enhance AI Verify’s capabilities.

In this video, we will walk you through the entire process, from architecture to customising your first plugin. This video is tailored for developers of all levels who want to enhance and contribute to the open-source community. Here’s what we will cover:

1. Introduction to AI Verify Architecture and Plugins
2. Customising Your First Plugin
   1. Testing Algorithm
   2. Display Widget

By the end of this video, you will have the skills to customise plugins to fit your unique requirements and the knowledge to contribute effectively to AI Verify.

The Model AI Governance Framework for Generative AI (MGF for GenAI) outlines 9 dimensions to create a trusted environment – one that enables end-users to use Generative AI confidently and safely, while allowing space for cutting-edge innovation. Recognising that no single intervention is enough to address existing and emerging AI risks, the framework offers a set of practical suggestions that apply as initial steps.

Read the primer on Project Moonshot to learn about our evaluation toolkit for Large Language Models.

AI Verify is an AI governance testing framework and software toolkit. AI Verify aims to help companies be more transparent about their AI systems, building trust through standardised tests and strengthening organisational processes.

“ISO/IEC 42001:2023 is a first of its kind international standards designed to ensure broad responsible adoption of AI. The novel approach, in conjunction with the family of standards that ISO/IEC JTC 1/SC 42 is developing, provides a portfolio of international standards that countries and regions can rely on to enable trustworthy and transparent AI.

This mapping between AI Verify Framework and ISO/IEC 42001 demonstrates Singapore’s strong support in advancing global harmonisation in a practical way. It also shows the growing number of countries that are leveraging ISO/IEC 42001 to enable trustworthy AI adoption.”

– Wael William Diab, Chair, ISO/IEC JTC1 / SC42 on AI standards

Both frameworks share a common goal in enabling organisations to strengthen their AI governance implementation. This crosswalk shows how the controls in ISO/IEC 42001:2023 are mapped to the process checks in AI Verify testing framework. Organisations can use AI Verify toolkit to strengthen their AI governance, and practically demonstrate alignment with ISO/IEC 42001:2023, without onerous cost.

We invite interested organisations to use AI Verify to meet the common outcomes of both frameworks and share your use case with us.

The Catalogue of Large Language Model (LLM) Evaluations sets out a comprehensive taxonomy that organises different domains of LLM evaluations to provide organisations a holistic overview of the available tests today.

It seeks to contribute to global discussions on safety standards by recommending a minimum baseline set of safety evaluations that LLM developers should conduct prior to LLM release.

The AI Verify Foundation welcomes initial comments and feedback on this draft release, which can be sent to info@aiverify.sg. We are currently establishing a more convenient method to receive and incorporate feedback from the community, and will update in due course.

This crosswalk will allow companies to use AI Verify to achieve the desired outcomes of both AI Verify testing framework and US NIST AI Risk Management Framework in promoting trustworthy and responsible AI.

The development of the crosswalk is an important step towards harmonisation of international AI governance frameworks to reduce industry’s cost to meet multiple requirements. The joint effort also signals Singapore’s and the US’ common goal of balancing AI innovation, maximising benefits of AI technology while mitigating technology risks.

We invite interested organisations to use AI Verify to meet the common outcomes of both frameworks and share your experience with us.

Read the primer of AI Verify to learn about our AI governance testing framework and software toolkit.

This paper from Singapore’s IMDA and Aicadium raises for discussion policy ideas on building an ecosystem for trusted and responsible adoption of AI, in a way that encourages a positive loop – to spur innovation and tap on opportunities afforded AI, ever more so with the advent of Generative AI.

By sharing ideas on practical pathways for governance, the paper seeks to enhance discourse and foster greater global collaboration to ensure AI is used in a safe and responsible manner, and that the most critical outcome – trust – is sustained.