Artificial intelligence (AI) and the Internet of Things (IoT) have quickly climbed in the ranking of the hottest terms in the technological arena. The integration of AI techniques within IoT is considered extremely advantageous to improve the quality of IoT applications. However, the introduction of intelligence into IoT raises new security challenges related to the additional complexity and the increased volume of smart services, especially when facing attacks from intelligent malicious IoT nodes. In spite of the urgent demand for security in AI-enhanced IoT systems, a complete solution involving trust management (TM) among IoT devices is missing. In this article, a four-phase TM approach assessing the trustworthiness of intelligent nodes is presented. This approach is capable of dynamically and diversely assessing the trustwor-thiness in each phase, namely access control, service provider selection, evaluation, and classification. The simulation results show the effectiveness of the proposed approach for securing intelligent IoT systems.