Generative AI is increasingly used informally in Air Traffic Management (ATM) for tasks such as flight‑plan generation, trajectory interpretation, and constraint checking. While it can reduce workload and speed up planning, its non‑deterministic outputs introduce safety and operational risks in human‑in‑the‑loop settings. This paper proposes the AI Trust and Assurance Layer (ATAL), a model‑agnostic decision‑assurance architecture that evaluates whether AI‑generated flight‑planning outputs are reliable enough for operational use. ATAL assesses outputs through three signals: (1) semantic stability under prompt variation, checking that core meaning remains unchanged; (2) operational consistency of structured outputs, verifying that waypoints, time windows and other fields are self‑consistent; (3) normative constraint validation against domain rules, ensuring compliance with altitude limits, airway restrictions and other hard constraints. These signals are mapped to a Decision Readiness Level (DRL) that helps human operators quickly decide whether to accept the result. An ATM‑inspired simulation study demonstrates how unsafe, inconsistent, or misleading AI outputs can be identified before they affect flight‑plan validation or execution. Although the case study focuses on aviation, the framework is transferable to other safety‑critical, regulator‑bound decision‑support domains that require human oversight.
Review: ATAL offers a systematic safety‑assessment pathway, laying the groundwork for controlled deployment of generative AI in high‑risk environments.