Adopting LLMs in Internet of Cloud Ecosystems: Identifying the Key Privacy Challenges

The rapid expansion of Large Language Models (LLMs) within Internet of Cloud (IoC) ecosystems creates significant risks regarding data privacy, security, and compliance. Additionally, although LLMs support real-time decision making and intelligent cloud services, their use within IoC ecosystems may expose sensitive data to privacy risks due to their complex design. This paper explores how ten of the most recognized privacy challenges such as: unauthorized data access, model inversion, and data leakage, arise during the deployment of 20 commonly used LLMs in IoC ecosystems. It begins by outlining each privacy challenge, then explains its specific impact on IoC ecosystems, regulatory compliance, and severity levels. Moreover, this study introduces a comparative matrix that evaluates each LLM’s level of compliance with these challenges. The matrix identifies which models meet privacy expectations and which do not, and includes examples of non-compliance, offering a clearer understanding of how these models differ in their exposure, vulnerability, and mitigation practices. The analysis reveals severe discrepancies across models, with many lacking sufficient transparency, effective consent management, and secure data deletion mechanisms. Finally, the findings emphasize the urgent need for a comprehensive privacy-by-design strategy and AI alignment protocols tailored to cloud-based LLM deployments.