Abstract: With the development of 6G non-terrestrial networks (NTNs) and space-air-ground integrated networks (SAGINs), communication coverage is expanding from terrestrial networks to aerial and spaceborne domains, while network openness, wide-area connectivity, and heterogeneous integration are becoming more pronounced. These trends improve global communication capability, but they also expose aerospace communications to serious security threats, including signal interception and platform exposure. As a result, covert communication has become an important means of protecting aerospace information security. At the same time, the rapid variations in network topology and channel conditions, the resource constraints of heterogeneous platforms, and the continuing evolution of warden capabilities make traditional covert communication methods less effective in complex aerospace environments, especially when those methods rely on strong prior knowledge and static optimization. This paper surveys AI-enabled aerospace covert communications. It reviews fundamental issues, including covertness metrics, theoretical limits, and extensions to aerospace scenarios, and examines the limitations of conventional covert communication methods in such settings. It also summarizes recent progress in applying artificial intelligence to warden-threat perception and state inference, resource coordination under covertness constraints, covert signal and camouflage jamming generation, and the design of anti-detection and covert transmission strategies. In addition, it discusses deployment strategies and representative application scenarios under the constraints of spaceborne and airborne heterogeneous computing platforms. The paper concludes by outlining the main challenges in this area and discussing future research directions.

Key words: aerospace covert communications, artificial intelligence, space-air-ground integrated networks, heterogeneous platform adaptation, intelligent covert communications