Recent geophysical and geotechnical data acquired on the Nice shelf to the east of the 1979 landslide source area, Suggest slow deformations processes which could lead to future catastrophic slope failure. According to these preliminary interpretations, it is of major interest to perform a slope stability evaluation to define the hazard and quantify the danger related to a probable instability on this slope. A probabilistic approach is proposed here using a modified version of the SAMU_3D model, a 3-D slope stability software recently developed by Sultan and others to account for complex geometry. The 3-D analysis is based on the upper bound theorem of plasticity developed by Chen and others. One of the main features of the original model is to allow complex critical failure surfaces, suitable for complex bathymetry (i.e. canyons). A probabilistic approach was added to the former deterministic model to consider the effect of sediment parameter variability and uncertainty (undrained shear strength and unit weight) on the likelihood of failure. Such an approach allows an estimation of the reliability of the results. Monte Carlo simulation was used to represent the variability of the factor of safety given a specific number of trials. Identification of the critical failure Surface previously based on a deterministic analysis is thus performed in terms of probability of failure (or probability of a factor of safety lower than a reference value). According to the undrained shear strength distribution profiles with depth, obtained using different models (down to 30 and 60 m depth) at several sites and to the parameter uncertainty, high probability of failure (around 50%) is found for the Nice slope indicating that the sediment in this area is highly metastable.