Diffusion at infinite dilution of U in metals, with particular emphasis in those used in nuclear facilities, is revisited. Early works present some particularities such as activation enthalpies lower than the vacancy formation enthalpy in the matrix, large differences with self-diffusion in the base material, up to four orders of magnitude differences between measurements performed by different authors in similar temperature ranges, etc. In particular U self-diffusion was qualified as abnormal when compared with other metals. Recent studies by means of α-spectrometry reveal a normal behaviour: activation enthalpies and pre-exponential factors similar to the self-diffusion one and diffusion coefficient values in the same order of magnitude than self-diffusion. The possible influence of short circuits, impurities and/or uncertainties in the techniques used in the early works is discussed in order to explain the differences obtained.