This study presents a fault-tolerant control framework for a six-degree-of-freedom aircraft aimed at preserving flight control system performance in the presence of software faults. As flight systems become increasingly dependent on software-based architectures, faults in the command generation path can lead to noticeable degradation in control quality. In the proposed structure, classical controllers are designed for the longitudinal and lateral channels, and a software fault is intentionally introduced in the lateral channel to evaluate the effectiveness of the method. To compensate for the failure effect on the aileron command, two data-driven models, namely a Gated Recurrent Unit (GRU) and a Temporal Convolutional Network (TCN), are trained using flight time-series data to generate auxiliary commands in parallel. The outputs of these two models, together with the output of the main controller, are fused through a robust voting mechanism. By relying on health indicators and reducing the influence of the faulty path, this mechanism provides a more stable final command to the system. To investigate the robustness of the proposed framework, the aircraft is subjected to severe atmospheric disturbances generated based on the standard Dryden turbulence model. Simulation results demonstrate that the proposed approach can mitigate the impact of the fault, preserve the continuity of lateral channel performance, and improve the stability of the system response under faulty and turbulent conditions without requiring an accurate fault model. Overall, the integration of classical control, data-driven models, and robust voting offers an effective solution for enhancing fault tolerance in flight control systems.