Stability analyses of any excavations within the rock mass require reliable geotechnical input parameters such as in situ stress field, rock mass strength and deformation modulus. These parameters are intrinsically uncertain and their precise values are never known, hence, their variability must be properly accounted for in the stability analyses. Traditional deterministic approaches do not quantitatively consider these uncertainties and variability in the input parameters. To incorporate these variability and uncertainties stochastic approaches are generally used. In this study, a stochastic assessment of pillar stability using Artificial Neural Network (ANN) is presented. The variability and uncertainty in the rock mass properties at the Laisvall mine were quantified and the probability density function of the deformation modulus of the rock mass was determined using probabilistic approach. The variability of the in situ stress was also considered. The random values of the deformation modulus and the horizontal in situ stresses were used as input parameters in the FLAC3D numerical simulations to determine the axial strain in the pillar. ANN model was developed to approximate an implicit relationship between the deformation modulus, horizontal in situ stresses and the axial strain occurring in pillar due to mining activities. The closed-form relationship generated from the trained ANN model, together with the maximum strain that the pillar can withstand was used to assess the stability of the pillar in terms of reliability index and probability of failure. The results from this study indicate that, the thickness of the overburden and pillar dimension have a substantial effect on the probability of failure and reliability index. Also shown is the significant influence of coefficient of variation (COV) of the random variables on the pillar stability. The approach presented in this study can be used to determine the optimal pillar dimensions based on the minimum acceptable risk of pillar failure