The swift adoption of Artificial Intelligence (AI) alongside the Internet of Things (IoT) has revolutionized smart cities, allowing for improved urban services such as traffic control, energy management, healthcare delivery, and public safety. However, the extensive generation and transmission of sensitive data by IoT devices present substantial cybersecurity risks. Unprotected data transmission can result in privacy violations, unauthorized access, and vulnerabilities within systems, jeopardizing the integrity and efficiency of smart city frameworks. This paper investigates the essential elements of secure data transmission within AI-driven IoT networks. It analyzes different encryption techniques, AI-enhanced Intrusion Detection Systems (IDSs), and decentralized frameworks based on blockchain to guarantee data integrity and confidentiality. Moreover, we emphasize the importance of federated learning, which enables distributed AI models to enhance their performance while keeping sensitive information localized, thereby reducing the likelihood of data breaches. Significant challenges are addressed, including the computational constraints of IoT devices, the diversity of IoT networks, and the requirement for low-latency communication in real-time scenarios. Innovative solutions, such as Quantum-resistant cryptography and the potential of 6G technology, are also examined. The paper concludes by outlining future research and development pathways aimed at improving the security, scalability, and efficiency of IoT networks within smart cities.