Singapore Proposes New Global Standard for Testing Generative AI Systems

Gate News message, April 20 — Singapore’s Infocomm Media Development Authority (IMDA) and Enterprise Singapore have proposed a new international standard for testing generative AI systems. The proposal will be presented at an ISO meeting in Singapore, the first such meeting held in ASEAN, bringing together more than 35 national bodies and 250 AI experts.

The initiative builds on Singapore’s earlier AI testing efforts, including the AI Verify toolkit and a starter kit for large language model applications. The AI Verify Foundation, a Singapore-backed organization with premier members including IMDA, Google, Microsoft, and IBM, has developed open-source tools for AI testing and governance. Its offerings include the AI Verify toolkit for governance testing and Project Moonshot, which combines benchmarking with red teaming methodologies.

The proposed standard is designed to complement ISO/IEC 42001, the international standard for AI management systems. It may help define technical requirements for compliance with regulations such as the European Union’s AI Act and California’s proposed SB-53 on AI safety and oversight. The framework covers quality traits including safety and employs methods such as red teaming for prompt-based text-to-text generative AI systems. This development could expand the market for AI assurance services, enabling specialist firms to help organizations meet emerging global benchmarks.