This study examines how the United States Air Force describes and delimits AI-enabled autonomy and human control in the development of Collaborative Combat Aircraft (CCA). It addresses a gap between broad research on human–autonomy teaming and the more limited knowledge of how a concrete military organization formulates these issues in an ongoing air power programme. The study is designed as a qualitative single-case study based on open-source strategic, policy-related and programme-near documents published between 2019 and 2025. Parasuraman, Sheridan and Wickens’ model of human interaction with automation is used as the analytical framework. The analysis shows that United States Air Force does not present CCA as a uniformly autonomous system. Instead, autonomy is distributed across different functions. Higher degrees of autonomy are associated mainly with information acquisition and, to some extent, information analysis, while decision-making and especially action implementation are described more cautiously. Human control remains central but is reformulated from direct control towards supervisory control and high-level direction. The study concludes that CCA is framed as varied autonomy under continued human control rather than as a move towards full autonomy.