It has recently been suggested that perhaps as many as 20% of all mitochondrial proteins are regulated through lysine acetylation while SIRT3 has been implicated as an important mitochondrial protein deacetylase. It is therefore of crucial importance that the mitochondrial localization of potential protein deacetylases is unambiguously established. Although mouse SIRT3 was recently shown to be mitochondrial, HsSIRT3 (human SIRT3) was reported to be both nuclear and mitochondrial and to relocate from the nucleus to the mitochondrion upon cellular stress. In the present study we show, using various HsSIRT3 expression constructs and a combination of immunofluorescence and careful subcellular fractionation, that in contrast with earlier reports HsSIRT3 is exclusively mitochondrial. We discuss possible experimental explanations for these discrepancies. In addition we suggest, on the basis of the analysis of public genome databases, that the full-length mouse SIRT3 protein is a 37 kDa mitochondrial precursor protein contrary to the previously suggested 29 kDa protein.