Singapore will launch Asia's first AI testing accreditation scheme in the third quarter of 2026 to certify firms that test and jailbreak AI systems for weaknesses [1, 2]. The program, called the AI Tester Accreditation Programme (AITAP), is led by the non-profit AI Verify Foundation under the Infocomm Media Development Authority (IMDA) [1, 2].
AITAP aims to provide a clear, unified assessment standard for AI testing firms to ensure trustworthy and reliable AI systems. Minister for Digital Development and Information Josephine Teo said the programme "aims to build a pipeline of competent testing firms that can rigorously assess AI systems across industries, from healthcare and finance to hiring and public services" [1].
AI testing includes simulating attempts to bypass AI safety rules, such as making chatbots generate malware or disclose confidential information [1]. Testers applying for accreditation will be assessed on technical competency, company standing, and financial sustainability following IMDA guidelines [1, 2].
So far, seven companies including EY, PwC, Advai, AIDX, and Knovel Engineering have expressed interest in applying for certification [2]. The scheme aligns with Singapore's risk-based and pragmatic approach to AI governance, which balances innovation with safety. Minister Yang Li Ming said, "安全不是为了对创新踩刹车，而是我们价值定位的重要组成部分。我们希望企业、公共机构和公民把AI深度融入重要系统。" [2]
The accreditation programme was announced on May 18, 2026, by Minister Teo at the International Scientific Exchange on AI Safety conference held in Singapore [1]. The 2026 AI Safety conference brought together over 100 AI scientists from 13 countries to discuss AI safety and prepare updates to the Singapore Consensus on AI governance [2].
Later in 2026, Singapore plans to update the Singapore Consensus to include new focus areas such as agent-based AI risk management and open-weight model security [2].
AITAP is expected to open for applications in Q3 2026, marking a concrete step to certify firms that test AI safety across key sectors in the region [1, 2].
