Industry 5.0 smart manufacturing environments integrate industrial Internet of Things (IIoT) devices, cyber-physical systems, cloud services, edge computing, intelligent automation, and human-centric production processes. This convergence increases operational intelligence but simultaneously expands the attack surface and introduces heterogeneous, dynamic, and distributed security requirements. Conventional intrusion detection systems (IDSs), particularly static rule-based and signature-oriented approaches, face limitations in recognizing evolving attacks, zero-day behaviors, distributed threats, and context-dependent anomalies. This research proposes an adaptive AI-driven intrusion detection framework for Industry 5.0 smart manufacturing environments by synthesizing machine learning, clustering, real-time detection, cloud-edge security, and threat-action analysis principles from the provided literature. The proposed methodology combines preprocessing, behavioral feature extraction, adaptive clustering, supervised anomaly classification, threat-context analysis, and continuous model updating. K-means-based behavioral grouping provides an initial mechanism for identifying deviations, while adaptive learning improves responsiveness to changing industrial traffic. The framework further incorporates cloud-edge considerations to reduce latency and support distributed security decisions. The analysis indicates that adaptive AI can improve the conceptual resilience, scalability, and responsiveness of intrusion detection compared with static approaches, although computational overhead, data quality, model drift, false positives, and deployment complexity remain significant limitations. The study positions adaptive intrusion detection as a foundational security mechanism for trustworthy Industry 5.0 manufacturing infrastructures.