AI in cities: safety gains without a surveillance creep

Civic AI that improves safety without creeping surveillance Cities can get real wins from prediction and monitoring - traffic signal timing, emergency lane detection, noise and air quality dashboards - without building an identity dragnet. The dividing line is purpose binding and proportionate retention. Governance that works Every dataset has a declared purpose and a retention window measured in days or weeks, not years. Operations are auditable by a third party, and residents have an appeals channel when systems affect them. Publishing aggregate dashboards builds trust and catches drift early. Face recognition for generalized tracking stays off the table. Procurement checklist Vendors must state data flows, retention, model types, and a clear redress path. Pilot on a small district first with metrics visible to residents before city‑wide rollout.