Toward Intelligent and Secure Satellite Communications: Agentic AI for Covert Semantics

With the rapid growth of satellite service, efficient and secure transmission in satellite networks has become a central challenge. To address it, semantic communication can improve the efficiency by conveying only the essential meaning of information, while covert communication can ensure the security by concealing the transmission behavior. However, existing solutions often treat these two objectives in isolation and lack intelligence for the joint optimization in a dynamic environment. In this article, we first provide a comprehensive overview of the integration of semantic communication and covert transmission in satellite networks, highlighting its representative applications and fundamental challenges. Then, we introduce the concept of large language model (LLM) enabled agentic AI with its capabilities and enabling technologies. Specifically, we propose an LLM-enabled agentic AI framework for covert semantic satellite communication to unify the semantic understanding, reinforcement learning (RL) based covert transmission and knowledge-driven reconstruction for intelligent and secure information delivery. We present a case study to demonstrate the effectiveness of this approach. Finally, open issues and future directions are discussed.