Urban Security and Technology: Surveillance (Lyon and Graham)

Technology-driven urban security is the use of cameras, sensors, drones, facial recognition and predictive algorithms to monitor and manage city space, and it has become one of the fastest and least debated transformations in contemporary urbanism. London has a camera for every few dozen residents, China has deployed hundreds of millions, and cities around the world buy systems that promise to prevent crime before it happens. Two authors have studied what that shift means: the Canadian sociologist David Lyon, founder of surveillance studies, and the British geographer Stephen Graham, who in Cities Under Siege (2010) traced the passage of military technology into urban policing.

David Lyon: surveillance as social sorting

Lyon, in Surveillance Studies: An Overview (2007) and The Culture of Surveillance (2018), proposes understanding surveillance not as a single panopticon but as social sorting: systems collect data to sort people into categories, suspicious, creditworthy, desirable, and treat them differently according to the category. That sorting is invisible to those subjected to it and tends to reproduce existing biases, because algorithms learn from police data that already over-represent poor and racialised neighbourhoods. Lyon adds that surveillance has become culture: people take part in it, share their location and images, and normalise it as a condition of urban life.

Stephen Graham: from war to urban policing

Graham shows where the tools come from. Facial recognition, drones, network analysis and prediction software were developed for the wars in Iraq and Afghanistan and for the control of the Palestinian territories, and returned to Western cities as security products sold by the same companies. His thesis of the new military urbanism holds that the city has been redefined as a battlefield and the citizen as a potential suspect, and that technological security concentrates on borders, shopping malls, financial districts and stigmatised neighbourhoods, that is, on the edges between rich and poor.

What the evidence says about cameras, prediction and facial recognition

The evidence on effectiveness is more modest than the market promises. Systematic reviews by Brandon Welsh and David Farrington of CCTV find significant crime reductions only in car parks and small or no effects on streets and public transport. Predictive policing programmes such as PredPol, used in Los Angeles and other cities, were dropped after showing that they sent patrols to the same neighbourhoods as always without reducing crime. And facial recognition fails more often with women and dark-skinned people, as Joy Buolamwini's 2018 Gender Shades study demonstrated, which has produced documented wrongful arrests.

Regulation: San Francisco, the EU AI Act and algorithm registers

Regulatory responses have begun. San Francisco banned municipal use of facial recognition in 2019 and dozens of American cities followed. The European Union passed in 2024 an artificial intelligence law that restricts real-time biometric identification in public spaces to exceptional cases with judicial authorisation. Barcelona and Amsterdam have created public registers of the algorithms they use and ethics committees. Lyon and Graham call for going further: evaluating every system by its proven effectiveness and its impact on rights before buying it, and returning to citizens control over the data the city collects about them.

The technological transformation of urban security raises, at bottom, the same question as security always has: for whom and against whom. A camera can protect a bus stop or watch a protest; an algorithm can optimise the ambulance or flag the neighbour. Lyon and Graham agree that the difference lies not in the technology but in who governs it, with what transparency and what limits, and that a city that buys security without asking ends up buying surveillance.