The increasing sophistication of multinational defense finance consortia requires an agile, privacy-preserving machine learning paradigm to handle distributed intelligence beyond sovereign boundaries. Classical centralized AI solutions are highly susceptible in high-risk defense areas, where financial, geospatial, or sensor telemetry might be disclosed and thus jeopardize national security. To solve this, we present a secure federated machine learning (FL)-based solution and introduce encrypted AI workflows for decision making in coalition defense finance. Our architecture allows the secure collaboration between the participants at the same time preserving the strictness of data sovereignty and confidentiality. Using homomorphic encryption, secure multi-party computation (SMPC) as well as differential privacy, the encrypted model updates are exchanged between federated nodes without any disclosure of the raw data. The system combines geopolotical intelligence (GEOINT) in real time, data from defense logistics sensors, and distributed financial ledgers from defense acquisition systems. In this way the model integrity and auditability is achieved in the architecture, by having zero- knowledge proofs and blockchain based consensus to provide tamper-evident model provenance and decentralized trust. Additionally, our approach employs transformer-based models and graph neural network (GNN) for federated settings to learn the latent defense-financial patterns between jurisdictions. Real-time adversarial attack, model poisoning and inference leakage detection modules are monitoring the system. We demonstrate the utility of our approach to the synthetic multinational defense scenario where trust and budgeted logistics are encrypted and inline with encrypted budgeting: we perform validation through a case study of encrypted coordination of logistics and budgeting among five allied states under stress due to cyber-warfare conditions. We present high predictive accuracy, model convergence stability, and robustness to attack vectors, while at the same time guaranteeing regulatory compliance with regional privacy laws. This is a landmark in secure, intelligent collaboration in multi-national defence finance, where AI, encryption, and geopolitics meet.