ISO/IEC 42119-8 Generative AI Testing Standard (Proposal)

On 20 April 2026, the 17th ISO/IEC JTC 1/SC 42 plenary opened in Singapore — the first time in ASEAN, co-organised by IMDA and Enterprise Singapore, with 35+ national bodies and 250+ AI experts participating. Singapore formally tabled ISO/IEC 42119-8, which, if adopted, will be the world's first international standard for testing generative AI systems.

Two core pillars: - Benchmarking — using shared datasets to measure AI performance, solving the comparability problem of "what to test and how to score" - Red Teaming — simulating attacks to surface hidden risks, standardising "how to find what's hidden"

The proposal builds on IMDA's domestic testing infrastructure: the AI Verify Toolkit, the Starter Kit for Testing of LLM-Based Applications, and the Global AI Assurance Sandbox. Changi Airport's February 2025 ISO/IEC 42001 AI Management System certification — the world's first for an airport — supplied a working precedent that AI governance can be externally audited.

IMDA CEO Ng Cher Pong (in post since November 2025), in his opening address, said: "Standards setting cannot move at a glacial pace" — or it risks being outpaced by AI itself. He also stressed that standards must be representative across sectors, cultures and languages, and that Southeast Asia — one of the world's most diverse regions — must be plugged into standards-making.

ISO standards typically take years from proposal to publication. But once a proposal is on the table, the framing for global discussion is set — which is precisely how Singapore translates 0.07% of the world's population into G7-tier AI governance influence.