This paper designs a data-driven model predictive controller for vehicle lateral stability control. The data-driven model is used to predict the future dynamics of the vehicle and is identified by the input-output data obtained from the high-fidelity Carsim model. The steering angle increment of the front wheels is used as a complete excitation signal to identify the system, which produces outputs of yaw rate and side slip angle. The constructed data-driven model was first compared with the 2-DOF model. Then, the input data was adjusted based on the comparison results to enhance the recognition effect, allowing the identified model to align more accurately with the Carsim model. The rate of change of the front wheel steering angle increment was limited to between -0.02 and 0.02 to ensure driver comfort. The expected trajectory is obtained from the yaw rate gain formula of the 2-DOF dynamic model. The results show that the controller is effective for maintaining lateral stability control of the Carsim vehicle and can accurately track the desired value. Additionally, when compared to the 2-DOF model, the data-driven model demonstrates a better control effect.