Why AI-powered city cameras are sounding new privacy alarms

For decades, cars dictated urban planning in the United States. Few could have predicted that they would one day also double as nodes for surveillance. In thousands of towns and cities across the U.S., automatic license plate readers have been installed at major intersections, bridges and highway off-ramps. These camera-based systems capture the license plate data of passing vehicles, along with images of the vehicle and time stamps. More recently, these systems are using artificial intelligence to create a vast, searchable database that can be integrated with other law enforcement data repositories. As a scholar of technology policy and data governance, I see the expansion of automatic license plate readers as a source of deep concern. It’s happening as government authorities are seeking ways to target immigrant and transgender communities, are already using AI to monitor protests, and are considering deploying AI systems for mass surveillance. Eyes on the road Using cameras to track license plates dates to the 1970s, when the U.K. was embroiled in a long-simmering conflict with the Irish Republican Army. The Met, London’s police force, developed a system that used closed-circuit television cameras to monitor and record the license plates of vehicles entering and exiting major roads. The system and its successors were seen as useful crime-fighting tools. Over the next two decades, they expanded to other cities in the U.K. and around the world. In 1998, U.S. Customs and Border Protection implemented this technology. By the 21st century, it had started appearing in cities across the U.S. There are different ways for a jurisdiction to implement these systems, but local governments usually sign contracts with private companies that provide the hardware and service. These companies often entice authorities with free trials of surveillance equipment and promises of free access to their data in ways that bypass local oversight laws. AI thrown into the mix Recently, AI has