🧠 Food for thought

Implications, context, and why it matters.

Singapore’s AI standards plan builds on years of joint work

• Singapore’s proposal builds on years of public work with global tech companies 1.
• The AI Verify Foundation, a Singapore-backed group, supported that effort. It builds open-source tools for AI testing and governance. Premier members include IMDA, Google, Microsoft, and IBM 2.
• Its tools include the AI Verify toolkit for AI governance testing, plus Project Moonshot. Project Moonshot mixes benchmarking with red teaming, which sit at the center of the proposed standard 2.
• The foundation and IMDA ran a Global AI Assurance Pilot with more than 30 organizations. They also launched the Global AI Assurance Sandbox, which links companies using generative AI applications with specialist testers. Both efforts are meant to inform future standards for technical testing of generative AI applications 3, 4.

The standard could influence AI rules and grow assurance services

• The proposed standard is meant to sit alongside ISO/IEC 42001, the international standard for AI management systems 5.
• Standards like this may help define the state of the art for meeting rules such as the European Union’s AI Act and California’s SB-53, a proposed state bill on AI safety and oversight 6.
• It offers a qualitative framework with clear objectives for reviewing prompt-based text-to-text systems that use generative AI. It also covers quality traits such as safety, with methods such as red teaming 7.
• This could help grow a market for AI assurance services, where specialist firms may help companies apply and check themselves against new global benchmarks 3.