Urban AI is increasingly embedded in the governance of security, mobility, and everyday life through sensors, cameras, and algorithmic systems promoted under smart city agendas. While these technologies promise efficiency and protection, their rapid normalisation raises tensions between security, democratic accountability and individual freedom. This commentary critically examines AI-enabled urban surveillance as a contemporary mode of statecraft, arguing that dominant safety narratives often obscure deeper processes of behavioural regulation, data extraction and power consolidation within urban infrastructures. Drawing on examples from both democratic and authoritarian contexts, it shows how AI-driven monitoring can reshape citizenship, amplify inequality through algorithmic bias and erode privacy and autonomy. To address these challenges, the study advances a multi-level analytical framework tracing how smart city paradigms materialise as surveillance infrastructures, generate social consequences, and produce democratic tensions that diverge into care-oriented or control-oriented governance pathways. The commentary concludes that safeguarding urban freedom requires treating urban AI governance as a political project grounded in transparency, participation, civil liberties and democratic oversight, rather than a purely technical exercise.