Singapore takes the lead this week in championing AI safety, playing host to more than 250 experts from around the world and presenting a ground-breaking international standard for the testing of Generative AI. 

In a statement issued Monday (20 April), the country’s Infocomm and Media Development Authority (IMDA) said the proposed new standard would guide organisations around the world in testing their GenAI systems to ensure trustworthiness.

The new standard – the world’s first in this space – centres around benchmarking and red teaming. 

The standard will be discussed at the ISO/IEC JTC 1/SC 42 plenary meeting in Singapore this week, co-organised by the IMDA and Enterprise Singapore (EnterpriseSG) and hosted in the ASEAN region for the first time. 

The bi-annual plenary gathers more than 35 national bodies and over 250 AI experts from around the world, including the U.S., U.K., China, Japan, Germany, France, and South Korea.

“With the rapid development and pervasiveness of AI across ecosystems, it is crucial that there are globally recognised AI standards to ensure AI is used in a reliable and safe way,” IMDA said. 

The Singapore standard “establishes an important framework for AI testing that enhances the reproducibility and comparability of results. This, in turn, will drive assurance and overall trust in AI systems and enable safer, more reliable adoption by AI deployers and users”, the authority added. 

The new standard – the ISO/IEC 42119-8 – builds on IMDA’s past work in developing domestic testing frameworks, including the AI Verify Toolkit and the Starter Kit for Testing of LLM-Based Applications for Safety and Reliability and in the nascent field of assurance through the Global AI Assurance Sandbox. 

IMDA and EnterpriseSG are also hosting a series of capacity building initiatives on the sidelines of the plenary meeting.