The incorporation of IoT technologies in smart cities offers revolutionary improvement in urban management but collectively collects massive amounts of sensitive data, which results in privacy issues. AI privacy risks are discussed in this paper, and AI-based privacy preservation approaches mitigate these risks in smart traffic systems, smart health, smart, safe cities, smart grids, and smart communication systems within smart city environments. The uses of different tactics like differential privacy, federated learning, homomorphic encryption, and blockchain-based architectures are examined with reference to their roles in protecting data and users’ aegis. Further, based on collected data, we discuss the existing key performance indicators, some potential scalability issues, and new trends such as quantum cryptography, which can serve as the basis for improving the privacy of smart cities in the future. Finally, the role and the trade-off between data usage and privacy are discussed and further work is encouraged to address upcoming challenges in privacy-preserving urban IoT systems.