[Purpose/Significance] Trusted data spaces, in supporting cross-actor data circulation and intelligent service provision, are confronted with problems such as heterogeneous data resources, opaque algorithmic models, and insufficient collaborative trust among participating actors. To address these issues, this study explores the translation mechanism of data-model collaboration and the process of trust reconstruction in trusted data spaces, aiming to provide theoretical support for secure data circulation and sustainable operation of trusted data spaces. [Method/Process] Actor-Network Theory (ANT) was introduced as the analytical perspective. Following the principle of generalized symmetry, this study incorporated human actors, including technology developers, platform operators, and policy makers, as well as non-human actors, including data resources, algorithmic models, and computing infrastructures, into the same analytical framework. On this basis, the key heterogeneous actors and black-box nodes in the process of data-model collaboration in trusted data spaces were identified. Furthermore, along the translation chain of problematization, interessement, enrollment, and mobilization, the study theoretically analyzed the translation mechanism of data-model collaboration and the process of trust reconstruction. [Result/Conclusion] This study identifies the black-box obstacles in data-model collaboration in trusted data spaces, which mainly include the lack of trust in data circulation, the opacity of algorithmic operation, and the ontological heterogeneity between data and models, and reveals the interactions among these three types of obstacles in the collaborative process. It constructs a translation chain consisting of problematization, interessement, enrollment, and mobilization and explains that data-model collaboration in trusted data spaces is not a simple coupling of data and models. Rather, it is a process in which heterogeneous actors form an executable and traceable collaborative order through continuous translation, thereby promoting trust reconstruction. Specifically, data resources, algorithmic models, computing infrastructure, and human actors do not naturally achieve coordination merely through technical connection. Instead, they need to gradually define shared problems, align interests, embed roles, and stabilize collaborative networks through repeated interaction and negotiation. In this process, trust is not only supported by technical mechanisms but also generated and reinforced through the coordinated operation of standards, interfaces, rules, responsibilities, and governance arrangements.