Landslides are extremely important geomorphic events which sculpt river basin ecosystems by eroding hillslopes and providing sediments to coastal areas. However, at the same time landslides are hazardous events for socio-ecological systems causing enormous biodiversity, economic, and life losses in developed and in development countries. We propose a statistical spatially-explicit model based on a maximum entropy principle model (MaxEnt) for the prediction of precipitation-triggered landslides at the year-scale. The model is based on landslide occurrences, precipitation patterns, and environmental covariates at the basin-scale. The model predicts the size-distribution and location of over 27,500 historical landslides for the Arno basin in Italy which is considered as a case study of precipitation-controlled basins. Future landslide patterns are predicted for the A1B and A2 climate precipitation ensembles from 2010 to 2100. The spatial distribution of landslides, their size, and their potential hazard is calculated. The potential landslide-hazard is strongly correlated with the variation of the 12- and 48-hour precipitation with return time of 10 years. We assume a homogenous damage function in order to provide an average estimate of the potential hazard of landslides. The potential landslide-hazard is determined by 4-parameters of the double-Pareto landslide-size distribution: scaling exponents and truncation points of scaling regimes. Thus the landslide-size distribution is an indicator of the geomorphic effectiveness of precipitation. We observed an increase in potential landslide-hazard in the dry period 2040-2100 due to the activation of small landslides in remnant sites of past big landslides. On average, as the climate gets wetter the probability of large landslides gets higher. For a +20 and -15 mm variation of the 12-hour precipitation in 2020 and 2100 respectively the potential hazard of landslide is predicted 90 and 20 times higher than in 2011. For the Arno, the A1B and A2 emission scenarios do not produce relevant differences in the predicted landslide patterns, supposedly due to the small scale of the basin with respect to the scale of variability of precipitation. The model is proposed as a valuable risk-assessment tool under climate change scenarios. Further development is needed for calculating heterogeneous damage functions based on real exposure and vulnerability as a function of predictions of socio-ecological systems for the landscape analyzed. Our landslide modeling and assessment of landslide hazard is potentially applicable to any river basin worldwide in which precipitation plays a key role in landscape evolution.