Abstract: The continuous development of major engineering systems toward larger scales, higher integration levels, and more severe service environments has posed increasing challenges to the safety assessment of engineering materials. Infrastructure systems, such as oil and gas transmission pipelines, nuclear power facilities, offshore structures, and transportation equipment, are often required to operate under complex conditions involving high temperature, high pressure, corrosive media, cyclic loading, and long service durations. Under these conditions, materials undergo progressive degradation processes governed by the coupling of mechanical, chemical, and microstructural factors, leading to cross-scale damage evolution and long-term performance deterioration. These characteristics substantially increase the uncertainty and difficulty associated with service-safety evaluation and lifetime prediction. Conventional safety assessment approaches are primarily based on physics-based mechanistic models or empirical formulations. While mechanistic models offer clear physical interpretations, they generally rely on idealized assumptions and simplified boundary conditions, which limit their applicability in capturing nonlinear damage accumulation and stochastic degradation under realistic service conditions. In contrast, data-driven methods, including statistical analysis and machine learning techniques, have demonstrated strong capabilities in extracting correlations from large datasets and performing efficient predictions. However, their effectiveness is often constrained by data availability, limited interpretability, and insufficient extrapolation capability when applied to complex or previously unobserved service scenarios. In recent years, data–mechanism integrated approaches have been increasingly investigated to address the limitations of single-model strategies. This paper presents a systematic review of recent advances in data–mechanism integration methods for the service safety assessment of major engineering materials. Typical service-induced failure modes, including corrosion, wear, fatigue, and fracture, are first summarized, with emphasis on their multifactor coupling characteristics and damage evolution mechanisms across different length scales. Representative engineering cases are discussed to illustrate the complexity of material degradation processes and the challenges faced by traditional evaluation methods. Commonly used assessment methodologies are then categorized and analyzed, including mechanistic models, statistical reliability approaches, and machine learning-based prediction methods. The advantages and limitations of each category are critically examined. Particular attention is given to three representative data–mechanism integration frameworks: tandem models, parallel models, and deeply integrated models. Tandem models employ data-driven techniques to identify or calibrate parameters within mechanistic models to improve predictive performance. Parallel models combine predictions from mechanistic and data-driven models to enhance robustness. Deep integration approaches, such as physics-informed neural networks and digital twin frameworks, explicitly incorporate physical laws into learning architectures to improve generalization and physical consistency. Existing experimental studies and engineering applications indicate that data–mechanism integrated methods can provide more reliable predictions of corrosion rates, fatigue life, and damage evolution trends under complex service conditions compared with conventional single-model approaches. These methods offer a promising pathway for improving the accuracy and credibility of service safety assessments. Finally, key challenges and future research directions are discussed, including multisource data integration, uncertainty quantification, model interpretability, and the establishment of standardized evaluation frameworks. This review provides a systematic reference for advancing service safety assessment methodologies and supports the reliable operation and lifecycle management of major engineering systems.