The use of Machine Learning (ML) is widely spread across a variety of interdisciplinary applications. However, its application in the development of safety-critical systems raises numerous challenges. The deployment of a safety-critical system requires its qualification or certification by regulatory authorities or by law. Currently, the deployment of ML components in the safety industry poses new challenges and questions that need to be addressed to ensure that their use is feasible. In particular, there are organizations that are researching this feasibility, experimenting with the use of ML components, or evolving standards for certification. The purpose of this chapter is to report on the exploration of the use of ML techniques in safety-critical systems: safety requirements that may preclude the use of these techniques, the identification of appropriate ML approaches, and the development of a neural network and embedded on a computer. For this purpose, the attitude control of the UPMSat-2 satellite is used as a case study to develop an experiment.