An adaptive sliding-mode torque control system, which incorporates the merits of both the sliding-mode control and the adaptive algorithm, is proposed. The sliding-mode controller is constructed by two integral surface functions. The uncertainty is formulated as a random variable around a constant mean, so that an adaptive mechanism is used to estimate the constant mean and the bound of the random variable. Moreover, an application of the proposed torque control to the position control of a motor is also presented. Some experiments verify the control theory and demonstrate the usefulness of the proposed control scheme.