So far, the smart city has mainly been a spectator
Smart city projects have attracted a great deal of attention in recent years. They have measured air quality, tracked parking spaces, visualised traffic flows and analysed energy consumption. A truly smart city could draw conclusions from all this data and independently impose bans, divert traffic or control the heating systems in public buildings – and if only the AI could manage to synchronise a school’s heating schedule with the school holidays.
In many cities, there are still boundaries of responsibility, manual decision-making and organisational inertia standing between insight and action. The city sees a great deal, but takes no action. Yet a city that collects ever more data but fails to translate that data into action merely highlights its own inability to cope. That’s all well and good. But it’s simply not enough.
Physical AI turns infrastructure into an actor
Physical AI, on the other hand, integrates sensor technology, AI, edge computing, digital twins, automation and physical systems. The key shift lies in the fact that intelligence moves closer to the physical infrastructure. Data is not merely collected centrally and analysed later. It is processed where responses are needed: at junctions, in railway stations, in buildings, in energy facilities, in control centres, at ports or in critical supply networks.
The World Economic Forum describes Physical AI as an approach that combines IoT and AI, but with a crucial shift in perspective: a traffic camera at a junction does not merely detect that there are many cars there and send hours of video footage to a cloud. Instead, a local computer situated directly at the junction analyses the situation. The system recognises: traffic jams, accidents, ambulances, pedestrian traffic and roadworks. It can then react more quickly and reprogramme all traffic lights in the vicinity so that traffic is diverted away from the accident site.
The potential applications range from traffic management and the monitoring of air and water quality to public safety, fire safety and weather response.
This means a traffic light becomes more than just a traffic light. It becomes part of an adaptive traffic environment. An electricity grid becomes more than just a supply infrastructure. It becomes a learning system for load, consumption and stability.
The utopia is networked. So is the dystopia.
That is precisely why the debate must not wait until the systems have long been in operation. Physical AI Cities can make cities safer, more efficient, more sustainable and more liveable.
But the same technical foundation can also tip in the opposite direction.
If cameras, sensors, movement data, traffic flows, building data, administrative data and security information are brought together without clear rules, this does not automatically result in a better city. What emerges, initially, is a powerful surveillance infrastructure.
But without governance, connectivity turns into surveillance.
Without transparency, security turns into mistrust.
Without a clear purpose, urban intelligence turns into surveillance.
Without democratic legitimacy, the smart city becomes a technological imposition.
This is the crux of the matter: the utopia of the connected city and the dystopia of the fully monitored city utilise the same technological building blocks. The difference lies not in the sensor technology, but in the governance. A free society also thrives on the luxury of being imperfect.
Cities are acquiring a digital nervous system
The city of the future no longer merely waits for reports. As a connected, adaptive organisation, it recognises patterns in advance and stays one step ahead of the situation. This, in turn, enables it to prioritise resources.
This is particularly crucial when it comes to mobility, energy, security and crisis management. A burst water pipe, a major event, a power cut, a storm, an accident on a main transport route or a disruption to public transport are not isolated incidents. They trigger a cascade of events. Traffic is diverted. Emergency services need routes. Buildings and neighbourhoods adapt to changes in usage. Logistics come under pressure. Communication channels are in overdrive. This is where Physical AI would be a godsend.
Autonomy without governance is not innovation, but a risk
Or is it a curse after all? Who is authorised to decide when traffic should be diverted? And from where? What data is used for this? Who is liable if a system sets the wrong priorities? How transparent must algorithmic decisions be? Which interventions must remain subject to human approval? How can discrimination, surveillance, security breaches and dependence on specific providers be prevented?
This debate is particularly unavoidable in urban areas. Cities are not test laboratories with citizens as involuntary beta users. Physical AI impacts living spaces. It affects mobility, safety, energy supply, administration, health, participation and public order. Anyone who focuses solely on maximum automation has failed to grasp the issue. The crucial question is not how much a city can automate, but which decisions it is permitted to automate.
Physical AI cities become part of critical infrastructure
As soon as Physical AI becomes embedded in so many areas of life, it itself becomes part of critical infrastructure. Then it is no longer a matter of smart pilot projects, but of resilience, cybersecurity, recovery, traceability, data sovereignty and operational responsibility.
An adaptive city must not collapse if a cloud service fails or a provider changes its interfaces. It needs sovereign architectures, open standards, data spaces and clear operational models. Not everything has to be European. But everything critical must remain controllable.
This is where Physical AI Cities become strategically important for local authorities, municipal utilities, transport operators, security agencies and infrastructure providers. Anyone modernising urban systems today is not merely deciding on new technology. They are determining the city’s future capacity to act.
Conclusion: The city may become smarter, but it must not become opaque
The wrong response to Physical AI would be a fear of technology. The second-worst would be a romanticised view of technology.
Cities should not seek to hold back this development. Rather, they should prepare themselves for the regulatory, technical and ethical challenges.
Regulatory means: clear responsibilities, clear restrictions on use, clear liability rules, and clear limits on surveillance and data linking.
Technically, this means: interoperable platforms, urban data spaces, secure interfaces, edge architectures, resilient operating models and auditable AI systems.
Ethically, this means: transparency, democratic control, proportionality, participation and ultimate human responsibility where decisions have a profound impact on people’s lives.
