AI surveillance is already here. Urban governance isn’t.
World Cup host cities were also AI surveillance labs. A new study finds 4 out of 5 cities deploy these tools or plan to, yet fewer than a third have ethics oversight. That should alarm everyone.
When the largest World Cup in history ended at MetLife Stadium, it marked the culmination of a summer in which millions of visitors moved through American host cities under the eyes of computerized systems most of them never noticed.
The federal government directed more than $1 billion to World Cup security, funding facial recognition cameras in and around stadiums, police drones over crowds, and AI-driven monitoring of transit hubs. More than 120 civil society organizations issued a travel advisory warning visitors about the screening they would face.
The tournament turned host cities into laboratories for urban artificial intelligence, and the experiments ran on tourists.
The two of us have conducted research on the use of electronic monitoring in urban areas, and we do not write to argue that cities should abandon these tools. Tourism contributed $10.9 trillion to the global economy in 2024, roughly 10% of global GDP, and safety increasingly decides where that money goes. Thirty-two percent of travelers now call safety the primary factor in choosing a destination, up from 23% in 2021.
AI can genuinely help cities deliver that safety. London’s transit authority uses smart video to spot trouble on crowded platforms. Singapore’s police fly drones that map crowd density at large events in real time. Used well, these systems find lost children, reroute crowds before crushes form, and speed emergency response.
The problem is the gap between deployment and governance.
In research we published with the University of Pennsylvania’s Institute for Urban Research in December, we found that 52% of surveyed cities already use AI in some capacity and another 32% plan to, yet only 31% have established a formal ethics review process — even though 88% of them rate ethics as very or absolutely important.
Cities are running the experiment first and writing the protocol later.
That order of operations should worry anyone who remembers that surveillance systems, once installed, rarely come down. Mega-events have long served as the justification for permanent expansions of monitoring capacity.
The gap matters because safety is not only physical. Our research frames tourist safety in three dimensions: physical safety from crime and disaster, digital safety from data breaches and fraud, and psychological safety, the felt sense of trust that makes a visitor relax in an unfamiliar place. An algorithm that marginally improves the first while quietly eroding the other two is a bad trade.
The public senses this. Two-thirds of respondents in surveys we reviewed insist that people should be informed when AI is used to prevent, detect, or investigate crime. London learned the hard way that deploying facial recognition vans in high-footfall areas without that transparency generates backlash that outlasts the deployment.
Some cities show what the alternative looks like.
Barcelona created an external advisory council of independent experts that evaluates high-stakes algorithms before deployment; that panel rejected fully automated decision-making outright. Helsinki maintains a public AI register that lists every municipal algorithmic system — what it does, what data it uses, and what safeguards constrain it — so any resident or visitor can look up exactly what is watching them. Neither city sacrificed safety for principle. They built the trust that makes safety technology legitimate.
American cities do not need to invent this from scratch. The road map we propose starts with three steps any city council can take this year:
adopt an AI governance framework tailored to tourism and public safety;
extend legal protection to residents’ and visitors’ data whenever private vendors handle it;
publish a public register of every AI system the city operates.
None of these steps requires new technology. All of them require political will.
The window is small. Nine in 10 mayors worldwide say they want AI in city management, while only 2% actively deploy it today.
That means the great wave of municipal AI adoption is still ahead of us, and so is the chance to shape it.
Los Angeles hosts the Olympics in 2028. The cities that spent this summer as laboratories should spend the next two years writing the rules — in public — before the next surge of visitors arrives. Tourists chose these cities for a sense of welcome and security. Cities owe them systems worthy of that trust.
Eugènie L. Birch is the Nussdorf professor of urban research at the Weitzman School of Design, and co-director, Penn Institute for Urban Research at the University of Pennsylvania. Mauricio Rodas is the former mayor of Quito, Ecuador, and professor of practice at the Weitzman School of Design. Luke Campo, a graduate of the University of Pennsylvania, supported this research.