During Asia Tech x Singapore (ATxSG) 2025, The Government of Singapore unveiled the draft Starter Kit for Safety Testing of Large Language Model (LLM)-Based Applications, aiming to provide the world’s first technical testing framework of real-world generative AI applications. The draft starter kit builds upon insights from the Global AI Assurance Pilot released earlier this year, which paired AI assurance and testing providers with generative AI deployers across 10 industries. By outlining best practices and detailed guidance, the draft provides actionable steps for organizations to address common challenges and risks, including hallucinations, undesirable content, data leakages, and vulnerability to adversarial prompts.
The draft starter kit builds upon Singapore’s efforts in advancing thought leadership in AI governance and practical tools to advance AI safety, including Project Moonshot, one of the world’s first LLM evaluation toolkits, and the Singapore Consensus on Global AI Safety Research Priorities. IMDA has also announced the completion of an additional crosswalk exercise to ensure interoperability with the U.S. NIST AI Risk Management Framework (RMF): Generative AI Profile.
