Cross-Cloud Data Privacy Enhancement: Optimizing Collaborative AI Systems through Federated Learning and LLMs

Abstract

With the increasing reliance on AI systems operating across multiple cloud infrastructures, ensuring data privacy while enabling efficient collaboration has become a critical challenge. This paper proposes a novel framework for Cross-Cloud Data Privacy Protection by integrating Federated Learning (FL) and Large Language Models (LLMs) to enhance the collaborative mechanisms of AI systems. The framework leverages FL to maintain data locality and preserve privacy, while LLMs act as intelligent coordinators, policy enforcers, and communication optimizers in the federated ecosystem. We present a modular architecture that addresses data heterogeneity, model coordination, and privacy threats. Simulations and case studies demonstrate the feasibility and performance advantages of the proposed approach in real-world scenarios such as cross-institutional healthcare systems. Our findings reveal that combining FL with LLMs can significantly improve security, trust, and operational efficiency in multi-cloud AI collaborations.