Artificial Neural Networks (ANN) are expected to be massively applied even in future safety critical systems where dependability - handling and reacting upon faults - is major concern. We briefly review how ANNs in system design relate to faults -- as a potential weakness and for mitigation. We provide some details on an ANN design for fault detection. The goal of the presentation is to raise the awareness for dependability in the thought-to-be-naturally-resilient ANN paradigm.