The fault tolerant tracking control problem for unknown nonlinear multi-input multi-output (ΜΙΜΟ) systems is investigated in this paper. A novel intelligent fault tolerant control (FTC) framework is proposed to solve the tracking control problem. To eliminate the effect of faults, a neural network adapted with the extended Kaiman filter (EKF) algorithm is created to online identify the unknown ΜΙΜΟ nonlinear systems, and then the steepest descent method and evolutionary programming (EP) algorithm is utilized to find a self-tuning PID controller for the adapted neural network. The resulted intelligent PID FTC tracker can not only achieve the tracking objective but also can maintain the stability and the expected performance when faults occur in system. Finally, a numerical example is given to illustrate the effectiveness of the proposed method.