A Decision-Assurance Layer for AI-Assisted Flight Planning

Summary

Generative AI is already being used informally in air traffic management for flight-plan generation, trajectory interpretation, and constraint checking. The paper proposes the AI Trust and Assurance Layer (ATAL), a model-agnostic architecture designed to determine whether an AI-generated flight-planning output is reliable enough for operational use. ATAL combines semantic stability under prompt changes, operational consistency of structured outputs, and validation against domain rules. It converts these signals into a Decision Readiness Level for human operators. An ATM-inspired experiment shows that the framework can identify unsafe, inconsistent, or misleading outputs before they influence flight-plan validation or execution. The authors present aviation as the demonstration setting and describe the approach as transferable to other safety-critical decision-support domains that require human oversight and regulatory constraints.