There has been numerous seismic hazard studies so far that includes Malaysian territories. However, there is a need to assess how reliable those studies are. Two main potential contributors to error have been identified: 1) seismic hazard analysis method and 2) ground motion prediction equation (GMPE). The amount of variation in predicting erroneous GMPE is huge. Thus, this paper concentrates on generating new GMPEs due to subduction specified for Malaysia and validated against developed GMPE. Empirical method for GMPE generation was utilized using recorded ground motion data acquired from the Malaysian Meteorological Department. The earthquakes were grouped according to the source, and only source types in the Sumatran subduction area were used due to the availability of enough data to identify a pattern. Three GMPEs were generated for three different source types, namely, shallow subduction earthquake, deep subduction earthquake, and backarc earthquake sources. Sumatran strike slip fault is considered within backarc seismicity. They were compared with the models proposed by Petersen (modified from Young), Atkinson and Boore, and Megawati. The comparison results showed that the proposed models are far superior at predicting the earthquakes in the Sumatran region, with percentage difference between estimates and the recorded values being the lowest. Therefore, the equations should be used in further seismic hazard analyses. Thus, this paper becomes a part of the recent initiatives in Malaysia to assess the hazards posed by earthquakes. © 2013 IEEE.