In this paper, the mean-square exponential stability is studied for neutral neural networks with time-varying delays, uncertain parameters and Lévy noise.Based on the sliding mode control, the generalized It? formula and the Lyapunov stability theory, several sufficient conditions are obtained for the reachability of the specified integral Sliding Mode Surface (SMS) and the mean-square exponential stability of neutral neural networks on SMS.Finally, the effectiveness of the obtained result is verified by a numerical simulation.