[en] Until now, the dominant factor in seismic hazard assessments has been historical seismicity. This approach is justified if the data derived from historical seismicity in a region are representative of its seismic activity and if we consider that in areas where damaging earthquakes have happened, they may happen again. This can be seen when seismotectonic relationships are well established in areas such as well localised plate boundaries (subduction zones, transform faults) and/or with a high rate of deformation. In these cases, the areas where future earthquakes might occur are usually well determined and the recurrences short enough on the time scale of historical observations. However, in areas where recent tectonic deformations are diffused and moderate, the historical period of seismicity data (a few centuries) is not long enough to observe a sample of historical seismicity representative of the present-day tectonic activity of the area. The studies on the most recent damaging earthquakes (El Asnam 1980, Spitak 1988, Cherchell Tipasa 1989) show that it would have been extremely difficult to predict the magnitude and the localisation of these events considering historical seismicity alone. It is the same in Provence (south of France) where the risk of occurrence of a damaging earthquake would have been underestimated before the June 1909 event. In all these cases, seismotectonic studies 'a posteriori' show that the geometry and kinematic of the faults responsible for these earthquakes can be identified and that all of them have presented seismic activity in the last thousand years. So it is necessary to emphasis a global approach of the problem through specific studies such as neotectonics, teledetection, geodesy, present-day stress field, strain field, paleoseismology etc. These studies will enlarge the period of observation compared with the data derived only from historical and present-day seismicity. In France seismicity is moderate and recent tectonic deformations are slight in comparison to neighbouring countries (Italy, Greece, Turkey). Only a few important earthquakes have been studied and the relationship between quakes and faults are still unknown. So it seems necessary to 'regionalise' the seismic hazard and to introduce concepts like 'seismotectonic domains' more than 'active faults'