Autonomy is moving from advisory roles to operational control in air and space systems where decisions occur at machine speed and failures can be fatal. We synthesize a practical definition of ethically aware AI in aerospace, outline accountability across the technical and governance stack, and present a four-layer architecture that connects value specification to model development, real-time decision execution, and audit & feedback. We provide assurance patterns (ethical governor, constraint solvers, and run-time assurance/Simplex), operational user-interface principles for meaningful human control, and three case studies: (A) LEO conjunction avoidance with auditable fuel/schedule trade-offs, (B) crewed depressurization with priority sequencing and overrides, and (C) ROE-gated escalation in contested airspace. We close with metrics and certification artifacts aligned to emerging trustworthy AI and certifiable-AI guidance (Abbreviations: ROE = Rules of Engagement; RTA = Run-Time Assurance).