Where Embodied AI Gets Proven: Singapore’s Opportunity in Public Safety

A robot can perform flawlessly in front of an audience and still not be ready for its first day of duty. The demo shows what a system can do when conditions are right. Deployment asks what it does when they are not: when the lighting fails, the layout is unfamiliar, or a person does something no one planned for.  

‍

That second question is where embodied AI succeeds or stalls, and public safety is one of the fields least able to get it wrong.

‍

Why Embodied AI Stalls between Demo and Deployment

Embodied artificial intelligence – AI that perceives and acts in the physical world through a body – tends to begin with bounded tasks and controlled conditions. The harder step is getting those capabilities to hold up amid the unpredictability of the real world.

‍

Public safety makes that challenge particularly acute because some of its environments actively work against the system. Contraband is hidden to avoid detection. Blast sites can shift underfoot. Chemical agents can degrade the very sensors a robot relies on. For a public safety officer, these conditions are the job. For a robot, they are what it has to be proven against.

‍

Testing Embodied Artificial Intelligence to Find Failure

Safety-critical fields such as aviation have long treated testing as a search for failure. They push systems past their limits before anyone depends on them.

‍

Embodied AI for public safety needs the same discipline: realistic mock-ups and trials designed to surface worst-case behaviour, with routine performance treated as the minimum. Benchmarks built for factory floors and staged demos can only tell us so much when failure carries operational consequences.  

‍

From Testing to Real Operations

FieldAI offers one example of what the next step can look like. Its Field Foundation Models provide general-purpose autonomy software for different types of robots, allowing them to operate without prior maps or fixed routes.

‍

FieldAI was part of the fifth cohort of Dimension X, Hatch’s open innovations challenge, where it developed and tested its technology against Home Team operational requirements. It has since formed a strategic partnership with Certis to advance scalable robotic applications in complex, live security environments, and opened a Singapore office to support ongoing deployments and integration work. The two companies are also working together on safety validation and operational frameworks.  

‍

The gap between promising technology and operational deployment is precisely what Hatch built Dimension X to help bridge. For embodied AI, both halves of that journey matter. Live environments expose systems to real-world variability, while purpose-built test environments can recreate situations that are too dangerous, rare or sensitive to learn from during routine operations.

‍

Singapore as a Proving Ground for Embodied AI

Singapore’s advantage lies in connecting operational problems with global technologies, validating those technologies against real requirements, and giving the strongest solutions a route into deployment.  

‍

Two kinds of evidence are now taking shape. One the public side, the new Home Team Humanoid Robotics Centre (H2RC) provides a dedicated testbed where Home Team departments, industry and academia can develop and test robotic capabilities. On the commercial side, private security operators such as Certis are bringing autonomous systems into live operations, showing how they perform outside controlled settings.

‍

Singapore has set out to have humanoid robots working alongside Home Team officers by 2028. The more consequential question may be what gets built around it. If Singapore can connect controlled validation with live operational experience, it can become the place where embodied AI earns its first day on duty.

‍

For Hatch, that is where the work lies: building the bridge between promising technology and proven public safety capability.  

‍

Building embodied AI, sensing or autonomy that could work in environments like these? Dimension X, Hatch’s global open innovation challenge, is open for applications from 12 October to 6 November 2026. Selected startups develop proof-of-concepts with Home Team end users, against real operational requirements. Apply for Dimension X today!

‍