The combat cloud is an integrated information and communication system essential for modern military operations. It connects all battlefield components, including drones, soldiers, vehicles, satellites, and Command and Control centers, into a cohesive real-time network. This architecture significantly enhances situational awareness, threat analysis, and decision-making processes. However, the increasing complexity of multinational operations necessitates solutions that ensure both tactical interoperability and data sovereignty. This paper introduces an ontological framework for the collaborative training of AI models through Federated Learning (FL), addressing integration challenges presented by traditional military standards like the JC3IEDM. While JC3IEDM offers a standardized vocabulary for defense systems, its complex structure often obstructs interoperability with domain-specific, lightweight ontologies typically used in AI applications. To bridge this gap, we propose an ontology translator that acts as a semantic bridge between a domain-specific FL ontology and a JC3IEDM-based ontology derived from its logical schema. This translator effectively resolves conceptual mismatches, facilitating accurate transformations of entities and attributes. By implementing our framework, we enhance semantic interoperability among diverse systems while reinforcing data sovereignty and boosting real-time decision-making capabilities. These elements are critical for the success of future multinational military operations in an increasingly complex global landscape.