It is common practice in atmospheric corrosion to fit the power-law function as a 'model' to help predict longer term corrosion loss or maximum pit depth. This function is also known as the 'log-log model' since it plots as a straight line on log-log paper. It used extensively including for aluminium alloys. A review of the relatively few data sets for long-term corrosion of aluminium shows systematic variation from the log-log power law model. Instead, in many cases the data can be interpreted as consistent with the multi-phase model previously proposed for steel corrosion in marine exposure conditions. Consideration is given to why the multi-phase model appears to be applicable also to long-term corrosion of aluminium alloys. This includes the important effect of corrosion products on oxygen diffusion, even though corrosion products of aluminium tend to occur in localized areas, that almost always involves pitting rather than general or uniform corrosion. It is proposed that the build-up of corrosion products forces a change to the chemical reactions responsible for aluminium corrosion, specifically forcing a change from cathodic oxidation and polarization in the shorter term to cathodic hydrogen reduction and polarization in the longer-term. The greater possible rate of outward diffusion of hydrogen compared to inward diffusion of oxygen is primarily responsible for the bi-modal character of corrosion loss or maximum pit depth as a function of exposure time.