Aim: This article examines the dual role of artificial intelligence (AI) in smart logistics, focusing on its capacity to strengthen human and operational safety while simultaneously creating new risks for data protection, cybersecurity and automated decision-making. Methods: The study applies a structured literature review covering publications from 2018–2025 in Web of Science Core Collection, Scopus, IEEE Xplore and Google Scholar, supplemented by comparative analysis of AI-enabled and conventional security approaches and a case-based synthesis of cyber-physical disruptions relevant to logistics. Results: The review indicates that AI can improve anomaly detection, video surveillance, network monitoring, predictive maintenance, routing and incident response. At the same time, the same dependence on data-intensive models creates vulnerabilities related to adversarial inputs, data poisoning, model theft, privacy leakage, algorithmic bias, model drift and service disruption. These risks can propagate from digital systems to physical logistics operations and therefore affect both data subjects and personnel. Conclusions: AI can provide reliable proactive support for smart-logistics security only when it is embedded in lifecycle governance that combines secure data management, continuous model validation, human oversight, resilient fallback procedures and compliance with relevant cybersecurity, data-protection and AI-governance requirements. The most appropriate approach is therefore not full automation, but risk-based human-in-the-loop deployment supported by auditable technical and organizational controls.