As the quest for performance confronts resource constraints, major breakthroughs in computing efficiency are expected to benefit from unconventional approaches and new models of computation such as brain-inspired computing. Beyond energy, the growing number of defects in physical substrates is becoming another major constraint that affects the design of computing devices and systems. Neural computing principles remain elusive, yet they are considered as the source of a promising paradigm to achieve fault-tolerant computation. Since the quest for fault tolerance can be translated into scalable and reliable computing systems, hardware design itself and the potential use of faulty circuits have motivated further the investigation on neural networks, which are potentially capable of absorbing some degrees of vulnerability based on their natural properties. In this paper, the fault tolerance properties of Self Organizing Maps (SOMs) are investigated. To asses the intrinsic fault tolerance and considering a general fully parallel digital implementations of SOM, we use the bit-flip fault model to inject faults in registers holding SOM weights. The distortion measure is used to evaluate performance on synthetic datasets and under different fault ratios. Additionally, we evaluate three passive techniques intended to enhance fault tolerance of SOM during training/learning under different scenarios.