This paper addresses the implementation of a data-driven control strategy in a real test bench based on proton exchange membrane fuel cells (PEMFCs). The proposed control scheme is based on unfalsified control, which allows adapting in real time the control law by evaluating the performance specifications based only on measured input-output data. This approach is especially suitable to deal with nonlinearity, model uncertainty, and also possible faults that may occur in PEMFCs. The control strategy has been applied to several experimental practical situations in order to evaluate not only the system performance, but also different fault scenarios. The experimental results have shown the effectiveness of the proposed approach to regulate the oxygen stoichiometry in real-time operation, as well as to maintain a proper system performance under fault situations. Also, a start-up mass-flow controller is added in order to bring the system toward its normal operating conditions.