Understanding a driver's behavior in a steering task is essential to the development of haptic guidance systems. This paper aims to predict driver torque control, especially when haptic guidance is part of haptic feedback. A new cybernetic driver model with an improved neuromuscular system is proposed and identified. It is assumed that the driver converts a target steering-wheel angle into torque by both indirect and direct control. Indirect control refers to the adaptation of the parameters of an internal model of steering compliance as perceived by the driver. Direct control accounts for the driver's corrective action through direct haptic feedback. The parameters of the model were identified with data collected from experiments conducted with a driving simulator. The results of identification were satisfactory and led to good representation of the driver's action, with or without haptic guidance. The model accurately predicted driver torque output. It can be used to study driver adaptation to haptic guidance systems.