Abstract Two sets of experiments on typical Class G well cement were carried out in the laboratory to understand better the potential processes involved in well leakage in the presence of CO2. In the first set, good-quality cement samples of permeability in the order of 0.1 μD (10−19 m2) were subjected to 90 days of flow through with CO2-saturated brine at conditions of pressure, temperature and water salinity characteristic of a typical geological sequestration zone. Cement permeability dropped rapidly at the beginning of the experiment and remained almost constant thereafter, most likely mainly as a result of CO2 exsolution from the saturated brine due to the pressure drop along the flow path which led to multi-phase flow, relative-permeability effects and the observed reduction in permeability. These processes are identical to those which would occur in the field as well if the cement sheath in the wellbore annulus is of good quality. The second set of experiments, carried out also at in situ conditions and using ethane rather than CO2 to eliminate any possible geochemical effects, assessed the effect of annular spaces between wellbore casing and cement, and of radial cracks in cement on the effective permeability of the casing-cement assemblage. The results show that, if both the cement and the bond are of good quality, the effective permeability of the assemblage is extremely low (in the order of 1 nD, or 10−21 m2). The presence of an annular gap and/or cracks in the order of 0.01–0.3 mm in aperture leads to a significant increase in effective permeability, which reaches values in the range of 0.1–1 mD (10−15 m2). The results of both sets of experiments suggest that good cement and good bonding with casing and the surrounding rock will likely constitute a good and reliable barrier to the upward flow of CO2 and/or CO2-saturated brine. The presence of mechanical defects such as gaps in bonding between the casing or the formation, or cracks in the cement annulus itself, leads to flow paths with significant effective permeability. This indicates that the external and internal interfaces of cements in wells would most probably constitute the main flow pathways for fluids leakage in wellbores, including both gaseous/supercritical phase CO2 and CO2-saturated brine.