AI agents could transform how cities respond to emergencies
A new AI platform helps departments collaborate faster during emergencies while protecting privacy, commercial confidentiality and operational security
Sponsored by City University of Hong Kong
Sponsored by
Sponsored by
Researchers at City University of Hong Kong (CityUHK) are harnessing the power of AI agents to help cities when disaster strikes.
“Urban emergencies require close cooperation between government departments and private organisations, each holding different data and expertise,” explains Jianping Wang, dean of the College of Computing at CityUHK. However, these stakeholders often struggle to collaborate quickly, especially in an emergency.
AI agents are one way to solve this problem, Wang says. “They can interpret a task, consult approved local data and use specialist tools to produce an assessment.” These agents could share authorised information with other agents and request relevant information from them.
An automated coordination process could save lives, but there are obstacles to realising it. “Effective coordination calls for secure ways to exchange relevant knowledge while respecting privacy, commercial confidentiality and operational security,” Wang says.
Wang and her CityUHK colleagues are developing an open ecosystem called SynapseCity, with the aim of creating the world’s first cross-departmental emergency coordination platform that also preserves privacy. The project received HKD 55.69 million (£5.35 million) from the Research Grants Council of Hong Kong.
“For Hong Kong, our aim is to achieve more efficient, effective, automated communication across departments during emergencies,” Wang says. “By enabling agents to exchange assessments and prepare response options together, we hope to reduce the delays caused by repeated manual communication and verification.” Organisations can develop their own agents and connect them to the platform or integrate existing agents and tools, she says.
Emergency responses are inherently interdisciplinary. “A single incident can involve the behaviour of a natural hazard, the stability of infrastructure, communications and the movement of people and resources,” Wang says.
Consequently, her team brings together researchers in AI, cybersecurity and communications with specialists in environmental hazards, civil engineering, energy systems, urban planning and optimisation. “These processes affect one another,” she explains. “A useful response system must account for knowledge from several disciplines throughout its design.”
AI can improve emergency responses in many ways, she says. AI agents facilitate automated cooperation, although data privacy constraints must be in place. “For example, a utility could communicate that a power interruption may affect drainage pumps without disclosing its underlying operational records,” Wang says. “Other agents could use that assessment to coordinate a response.”
In an unfolding emergency, AI agents can process information faster than a group of people spread across different departments or organisations. “Agents could bring together weather assessments, infrastructure conditions and transport updates, identify connections and flag inconsistencies or missing evidence,” Wang says. For example, during a rainstorm that could help responders understand how flooding and road closures jointly affect rescue access.
Using agentic AI to gather information could also reduce the burden on human decision-makers, she notes. “By handling routine information checks, prioritising urgent updates and comparing response options, AI agents could reduce the mental workload and help people focus on judgments that require their experience,” Wang says. However, the final decisions still sit with humans, she says.