Scope of the current study is the investigation of the possibility to use Artificial Neural Networks (ANNs) in the design of offshore pipelines against geohazards. In particular, the prediction of the behaviour of subsea pipelines that cross active seismic faults has been studied to improve a smart GIS-based tool that had been developed in the past. Initially, 2D finite element (FE) models were created. In the first model the fault rupture was simulated, while in the second model the impact of the imposed permanent ground displacements (PGDs) was investigated. Numerous parametric analyses were performed, that were emerged through the alteration of the sediment thickness (H), the angle of the internal friction (?) and the cohesion (c). For each and every one of the models several analyses took place and results were yielded regarding the PGDs of the seabed (U1, U2) and the pipe’s strain. Consequently, the results of the analysis were used for the development of the ???s. The H, ? and c parameters were the input variables and the U1, U2 and ? parameters were the output variables. Training of the ANNs took place with three training algorithms. Their ability to predict the results was investigated and error testing took place. Further testing was carried out through the execution of additional simulations, that they were not included in the original training data. The results were compared with the corresponding data of the neural network for method testing. The prediction error for the pipe's strain was smaller than 1%. In conclusion, the objectives were met in a sufficient degree. It was proved that ANNS can predict the behaviour of subsea pipelines, which are subjected to the displacements of faults’ rupture, provided that they are based on datasets that are extracted from accurate FE simulations.