Singapore's Global AI Assurance Pilot — 16 Testers Paired With 17 GenAI Deployers Across 10 Industries

In February 2025 Singapore's Infocomm Media Development Authority (IMDA) and the AI Verify Foundation launched the Global AI Assurance Pilot to explore and establish technical testing practices for generative-AI applications used in real deployments.
The pilot paired 16 specialist AI testing firms with 17 deployers of GenAI applications across 10 industries (including finance, healthcare and the public sector) and 9 geographies. Of the applications, 12 were internal-facing and 5 external-facing, and 12 kept a human in the loop.
The main report (June 2025) found that deployers were most concerned with accuracy and robustness, use-case-specific regulation and compliance, and content safety. Off-the-shelf LLM benchmark datasets were rarely used; use-case-specific test data was the norm. Human review and LLM-as-judge were the two most popular evaluation methods, the latter needing careful design and human calibration.
Findings fed a Testing Starter Kit for GenAI applications and a follow-on Global AI Assurance Sandbox. Limits: results are descriptive practitioner insights from 17 cases, and testing approaches depend heavily on context; no outcome or public-value metrics are reported.

Read the full analysis: https://aiverifyfoundation.sg/ai-assurance/2025-main-report/executive-summary/