Singapore proposes new global standard for AI benchmarking, red teaming

Singapore – Singapore has proposed a new international standard aimed at bringing consistency to how generative artificial intelligence systems are tested, as global efforts to govern the technology gather pace.

The proposed standard, ISO/IEC 42119-8, focuses on benchmarking and red teaming methodologies—two areas seen as critical to assessing the performance, safety, and reliability of generative AI systems. 

It is set to be discussed at the 17th ISO/IEC JTC 1/SC 42 plenary meeting, taking place in Singapore from April 20 to 24.

Co-organised by Infocomm Media Development Authority and Enterprise Singapore, the event marks the first time the bi-annual gathering is hosted in the ASEAN region. 

It will bring together more than 250 experts and representatives from over 35 national bodies, including the United States, the United Kingdom, China, Japan, Germany, France, and South Korea.

The move comes as governments and industry players race to establish guardrails for AI deployment. With adoption accelerating across sectors, standardised testing frameworks are increasingly viewed as essential to ensure systems behave as intended.

ISO/IEC 42119-8 aims to introduce consistent approaches to evaluating generative AI, improving the reproducibility and comparability of results.

As highlighted by Infocomm Media Development Authority Chief Executive Ng Cher Pong, “These are critical standards that will form the foundation of more trustworthy testing and facilitate more reproducible and comparable results.” 

By doing so, it seeks to strengthen trust in AI systems and support safer deployment across markets.

The proposal builds on Singapore’s earlier work in AI assurance, including the AI Verify Toolkit and testing starter kits for large language model applications. These initiatives have focused on creating practical methods for assessing AI systems in real-world scenarios, feeding into broader international standardisation efforts.

Beyond the standard itself, Singapore is also hosting a series of capacity-building initiatives alongside the plenary. 

These include training workshops on AI standards in partnership with the American National Standards Institute, policy-focused sessions for national bodies, and industry discussions under the AI Assurance Exchange led by the AI Verify Foundation.

The effort forms part of Singapore’s wider push to position itself as a hub for trusted AI development. 

Through initiatives such as its AI Safety Institute and participation in ASEAN-level governance frameworks, the country is seeking to shape how AI is tested, governed, and deployed across the region.

As generative AI continues to scale, the emphasis is shifting from innovation to accountability. Singapore’s latest proposal signals a growing push to define not just what AI can do—but how it should be measured.

The push also aligns with Singapore’s broader AI ambitions. Earlier this year, the government announced plans to establish a National AI Council chaired by Prime Minister Lawrence Wong to drive large-scale AI deployment across sectors such as manufacturing, finance, and healthcare.