As AI increasingly transforms SCM into proactive and automated environments, it simultaneously introduces complex cybersecurity vulnerabilities and stringent legal obligations under frameworks such as the EU AIA and the GDPR. This thesis explores the critical intersection of AI, trustworthiness, and supply chain operations, highlighting the necessity of ensuring systems are transparent, robust, secure, and accountable. To safely integrate these technologies into critical logistics infrastructures, the research proposes a Secure SCM-MLOps framework that synchronizes the core operational stages of the supply chain with the AI development, quality assurance, and production lifecycles. This integration is supported by a comprehensive five-layer governance architecture that operationalizes legal mandates into technical controls, emphasizing continuous monitoring, zero trust principles, and essential HITL oversight. The practical validity of this framework is evaluated through a scenario analysis involving ready mix concrete logistics, a highly sensitive environment where AI prediction failures regarding delivery times can cause severe structural damage known as cold joints. Utilizing the FAITH project Trust Guard tool, the study reveals that static vulnerability assessments are insufficient for securing such high stakes industrial operations against dynamic threats like data poisoning and evasion attacks. Furthermore, technical simulations conducted with the IBM ART on the TabPFN-2.5 foundation model provide empirical proof of these vulnerabilities, demonstrating that models deployed without proper data standardization suffer catastrophic accuracy drops, falling to zero percent when subjected to adversarial boundary attacks. The findings conclude that achieving defensive reliability and operational safety in smart supply chains requires moving beyond basic algorithmic implementations to a holistic approach that combines rigorous mathematical feature standardization within context adaptation and continuous human-centric governance.