Singapore to certify private sector AI testing firms

Singapore has announced a structured system to allow accredited testing firms to verify the integrity of an organisation’s artificial intelligence (AI) networks. 

In the third quarter of this year, the not-for-profit AI Verify Foundation (AVF), a subsidiary of the Infocomm Media Development Authority (IMDA), will launch the AI Tester Accreditation Programme (AI TAP).  

AI TAP would build an accredited list of third-party testing firms, which have the technical know-how, to rigorously assess AI systems across industries, from healthcare and finance to hiring and public services.  

This would be a first in Asia. 

Announcing this on Monday at the two-day International Scientific Exchange on AI Safety (ISE) 2026 conference, Singapore’s Minister for Digital Development and Information, Josephine Teo, said the scheme to accredit private sector AI testers was intended to set standards and to uphold a certain level of confidence in the testers themselves.  

The foundation noted that this would provide businesses with greater confidence in the assessments made by the accredited testing firms and, in turn, assure end-users of the reliability of the AI systems they were interacting with.  

“This [approach] was aligned to Singapore’s practical, risk-based approach to AI that maximises innovation and adoption with appropriate guardrails. These critical elements are needed to build a trusted ecosystem,” AVF said in a statement. 

The programme anchored accreditation to best practices, ensuring that testing service providers were assessed against a “meaningful and practical bar”.  

AVF noted that accredited providers would be recognised for capability in technical testing for AI risks and have the opportunity to connect with potential customers.  

The foundation noted that organisations would be able to build stakeholder confidence when their AI has been tested independently.  

At the same time, they could learn from experts on how to strengthen internal capabilities in AI testing. 

The AI TAP programme followed the success of the Global AI Assurance Sandbox that aimed to codify emerging norms and best practices around technical testing of Generative AI (Gen AI) applications.  

It looked at two elements of testing – what to test and how to test. 

Since its launch in February 2025, the Sandbox has tested 30 AI applications from 14 sectors, covering new archetypes such as agentic AI and risks such as vulnerability to prompt injections. 

AI scientists meet in Singapore 

The ISE event has brought together more than 100 leading AI scientists, researchers and AI labs from 13 countries for two days of dialogue on global AI safety research. 

Organised by IMDA and supported by the Future of Life Institute, a non-profit organisation, and Concordia AI, a social enterprise, it was one of the few global events focused on AI safety research. 

This was the second time Singapore was hosting the ISE, following the inaugural convening in April 2025. 

The inaugural conference last year established the Singapore Consensus on Global AI Safety Research Priorities, which has emerged as a significant milestone in uniting global experts to bridge the gap between science and policy. 

During the current ISE 2026 conference, an updated version of the consensus has been worked on and was scheduled for release in the second half of this year to address rapidly evolving technological threats to AI.

The upcoming revision would elevate societal resilience as a vital fourth pillar to harden public systems against AI failures and misuse to join the three existing areas of risk assessment, development and deployment, and post-deployment control.  

The revised version would also introduce targeted frameworks for agentic AI risk management amid rising AI autonomy, and stronger technical safeguards for open-weight models that were nearing frontier AI model capabilities.