Nowadays, robots are increasingly used in various surgery applications. Meanwhile, in many of these applications, the operation tools of surgical robots need to be in contact with human soft tissues. Furthermore, this tool-soft tissue interaction brings great challenges to robot control and system performance due to the nonlinearity, viscoelasticity, and uncertainties of the soft environment. To address these challenges and achieve the desired interaction behavior, a learning-based force controller for a surgical robot, which consists of a feedforward plus feedback controller, a radial basis function neural network-based controller, and an adaptive proportional-integral-type sliding mode control-based compensator, is presented in this letter. To display the stability of the proposed controller, the control system of the robot is analyzed through the Lyapunov method. Finally, several experiments are carried out in the robot prototype and the results illustrate that good tracking performance and guaranteed robustness can be obtained by the proposed controller.