Existing data on the release of fission products (FPs) from UO2 above 1000°C show that the dominant transport process consists of elementary diffusion within grains. For many FPs, the noble gases among them forming an exception, this diffusion is characterized by an activation energy of ∼2.6 eV, which is close to the one for oxygen and very different from the one for uranium. Assuming that oxygen diffusion represents the diffusion of FPs, it can be predicted that diffusion is enhanced when there is excess oxygen in the lattice. An empirical relation between the pertinent activation energy and the overstoichiometry induced by uranium fission (burnup) is given. The transport by diffusion has to be driven by some gradient, and it is argued that the temperature gradient dominates over the concentration gradient. This argument leads to a complete description of the release rate in terms of the grain size, the central and surface temperatures, and the heat of transport. The heat of transport plays a crucial role as it varies greatly for the various FPs. Existing data allow estimation of values ranging from 0.1 eV for refractory products to more than 100 eV for volatile products. These variations appear to be correlated with variations in the bond strengths between FPs and oxygen, being the more reactive element in UO2. An empirical model of the dependence of the heat of transport on this bond strength is given, so that release rates for all the FPs can be derived from chemical tables. Finally, consistency of the measured release data with other independently obtained fuel parameters is proven.