The quiet builders of the AI age: why Singapore is betting big on standards

Imagine you’re building a high-performance aircraft. You’ve got the engine, the wings, the sleek design. But without a safe, well-built runway, that plane is going nowhere. That, in a nutshell, is the relationship between cutting-edge AI and the often-unseen world of standards.

Recently, Singapore played host to a massive gathering of the people who build those runways. We’re talking about the 17th ISO/IEC JTC 1 SC 42 Meeting, the global brain trust for AI standardisation. Over a few days, delegates from more than 35 countries (and many more online) huddled together to do the unglamorous but vital work of agreeing on a common language for AI.

Why does this matter? Because standards are the quiet infrastructure that enables things to work together at scale. They’re the reason your phone can connect to a network in another country. And when it comes to AI, they’re the launchpad for rapid adoption.

The AI Management System is already a thing

Take a standard called ISO/IEC 42001 on AI Management Systems. Since it launched, it’s been translated into local standards in multiple places, including Singapore. More importantly, companies are actually lining up to get certified. The Changi Airport Group became the first in Singapore to get the nod, using it to bake clearer accountability and risk checks into every AI project across their organisation. It’s a prime example of how a standard can turn big, fuzzy principles into concrete, day-to-day controls. (We covered a story of this: MAS Proposes Guidelines on AI Risk Management: A Closer Look at Key Aspects – sahafiun reactions & reviews)

And the work is accelerating. The number of AI standards either published or in the pipeline has grown significantly in a short time. That’s necessary, given the breakneck speed of AI development. In just over three years, we’ve seen AI go from generative to multi-modal, and now to agentic AI. The standards work simply has to keep pace.

The three big challenges

Singapore’s leaders, who helped host the event, laid out three key themes where the community needs to focus its energy.

First, speed vs. consensus. Building agreement takes time, but AI waits for no one. The standards process can’t move at a glacial pace, or it risks becoming irrelevant. The solution isn’t simple, but it starts with having honest conversations about the gaps and opportunities.

Second, standards have to be for everyone. They need to be representative and inclusive, not just across different industries, but across languages and cultures. Southeast Asia is one of the most diverse regions in the world, so it’s crucial that the region is plugged into the process, not just a consumer of the output. A capacity-building workshop for ASEAN member states was held the week before the plenary, which is a step in the right direction.

Third, testing and assurance. A standard is just words on a page unless you can test against it. Good testing builds trust and encourages more widespread use. Here’s the big one: Singapore has put forward a new international standard, ISO/IEC 42119-8, to standardise testing methodology for Generative AI systems, aimed at strengthening the foundation for trustworthy AI testing. This is the first international standard of its kind for the testing of Generative AI systems. Singapore experts are also active in other areas like AI red teaming. To bridge the gap between standards and real-world practice, the AI Verify Foundation launched a sandbox to do testing on actual problem statements.

From paper to practice

Developing the standard is just the starting point. The real impact comes from adoption, and adoption depends on credible testing and certification. Singapore launched a pilot accreditation programme alongside their adoption of the 42001 standards. The idea was simple: when the standard was ready, the system to certify against it was also ready. This parallel approach helped accelerate adoption and reinforce credibility. And it’s not just big players; a startup using AI to digitally fingerprint food ingredients is also working towards certification, showing that accredited certification is becoming a competitive advantage, not just a compliance box to tick.

The value of standards isn’t an end in itself. Their real worth lies in how they are put into action to solve problems and enhance trust. As the global community works on these important tasks, the hope is that the conversations happening in meeting rooms spill over into coffee chats and corridor chats, sparking new ideas and projects. Because in the fast-moving world of AI, no one has a monopoly on what good looks like.