Singapore proposes global GenAI testing standard

Benchmarking and red teaming guide under review.

The government has proposed a new international standard for testing generative artificial intelligence systems to strengthen trust and consistency in how such models are assessed.

The proposed standard, ISO/IEC 42119-8, focuses on benchmarking and red-teaming methodologies for generative AI systems and is intended to improve the reproducibility and comparability of testing results, according to the Infocomm Media Development Authority and Enterprise Singapore.

The proposal will be discussed at the 17th ISO/IEC JTC 1/SC 42 plenary meeting in Singapore from 20 to 24 April, which will be held in ASEAN for the first time.

The event is expected to gather more than 35 national bodies and over 250 AI experts from countries including the US, UK, China, Japan, Germany, France, and South Korea.

IMDA said the proposed framework builds on Singapore’s earlier work on AI assurance, including the AI Verify Toolkit, the Starter Kit for Testing of LLM-Based Applications for Safety and Reliability, and the Global AI Assurance Sandbox.