A detailed stochastic model of HIV infection and AIDS for large cities in central Africa is described, which reproduces past events in Kinshasa, Zaire and projects rapid future spread of the disease, consistent with recent findings for Nairobi, Kenya. Most of the parameters used describe the behaviour of individuals, and it is therefore possible to look at the effects of changes in such behaviour, and thus to test various strategies aimed at providing effective public health policies. The model demonstrates that, if the spread of infection is to be controlled, changes in the behaviour of the major risk groups are essential. With appropriate modifications, this model could be adapted for use elsewhere in Africa.