Since 1996, more than 11 Mt of CO2 have been injected into a deep saline aquifer, the Utsira Sand formation, at the Norwegian Sleipner field. An unexpected application of the extensive seismic monitoring program over this field leads to the estimation of the depth dependence of the permeability anisotropy (strength and direction). Time-lapse seismic monitoring is used to follow the displacement of the injected CO2, considered as a permeability tracer. The upper half of the Utsira sand formation exhibits large anisotropy, with a ratio ζ between the maximum and minimum horizontal permeabilities larger than six. In contrast, ζ can be as small as two in the lower half. The direction of the maximum horizontal permeability does not exhibit substantial depth dependence and lies between N18°E and N34°E. This is in agreement with previous authors who pointed out a clear CO2 plume structure markedly elongated in the north-northeast–south-southwest direction. The small topographical trap at the top of the Utsira sand formation is a minor extrinsic cause of the measured permeability anisotropy, compared to the intrinsic effect of the formation permeability. This permeability information is crucial for reservoir simulation and for forecasting of the CO2 plume expansion for different scenarios of injection.