Aircraft loss of control in-flight (LOC-I) is a primary contributor to fatal accidents worldwide. As air traffic increases, current aviation control systems may be unable to adequately manage severe LOC-I related issues. In this paper, a data-driven system health monitoring (SHM) technique using an autoencoder (AE) is proposed to detect aircraft LOC-I precursors in real-time and to provide aircraft system level information to air traffic controllers (ATCs) for proactive aviation safety management. An air traffic simulator is utilized to investigate aircraft flight operations and trajectories based on flight phases and flight plans. To estimate nominal flight operations, an AE model is adopted. A statistical detection baseline is defined using multivariate Gaussian distribution to detect LOC-I precursors, which are statistically uncommon operation patterns. The proposed technique is validated using a case of LOC-I scenario. The novelty of this study lies in development of a real-time, data-driven LOC-I precursor detection technique, and an interface that can connect aircraft system health information with ATCs.