Field note 03 / SingaporeWhere trust meets deployment.
28 September 2026Public sourcesTrusted AI
A closer look at the public infrastructure Singapore is building around testing, deploying and governing AI systems in real settings.
Four signals.
Each row names a public signal, the structure it reveals and the question that remains open.
National priorities now frame AI as deployment and public good.
Smart Nation Singapore's 2026 update to the National AI Strategy describes refreshed priorities under a National AI Council and continues to frame Singapore's AI agenda around public good, adoption, capability and trusted development.
Read the National AI Strategy update ↗Why it mattersA strategy is most useful when it creates a common direction for researchers, firms and public institutions. Direction alone does not show how trade-offs will be handled in a specific deployment.
02 / TestbedPhysical AI is being connected to a real operating environment.
In May 2026, IMDA described collaborations involving government bodies, the Singapore Institute of Technology and eight industry leaders at Punggol Digital District to research, test and deploy physical AI. The same announcement positioned high-trust sectors as a focus for reliable and responsible deployment.
Read IMDA's deployment announcement ↗Why it mattersA test environment makes it possible to study a system in context. It also raises practical questions about oversight, affected people and the threshold for moving from a trial to normal operations.
03 / EvidenceThe agent sandbox treats behaviour in practice as a source of governance knowledge.
IMDA describes an AI Agents Sandbox launched with CSA, GovTech and Google in August 2025 to examine how computer-use agents behave in practice, including risks and unintended consequences. The work was translated into a public whitepaper and governance learning.
Read the AI Agents Sandbox insights ↗Why it mattersGovernance becomes more concrete when it learns from a system's observable behaviour. Public lessons are useful only if they remain specific enough to guide future design and accountability.
04 / OwnershipApplied AI programmes put the organisation's problem at the entry point.
AI Singapore's 100 Experiments programme evaluates whether a proposed solution can create substantial value and whether the organisation has an appropriate Singapore-based technical team to work on it. It is a problem-led route to applied AI work.
Read about AI Singapore's 100E programme ↗Why it mattersA deployment has a better chance of becoming accountable when an organisation owns a real problem and has the capacity to participate in the work, rather than simply acquiring a tool.
SUPERINTELLIGENCE SG