Neutralizing antibodies are likely to play a crucial role in a preventative HIV-1 vaccine. Although efforts to elicit broadly cross-neutralising (BCN) antibodies by vaccination have been unsuccessful, a minority of individuals naturally develops these antibodies after many years. Thus understanding how these antibodies evolve could provide a template for HIV vaccine design. To date we have studied ∼80 HIV-infected individuals in the long-term CAPRISA cohort and found that approximately a quarter develop cross-neutralising antibodies by 3 years of infection. Analysis of longitudinal samples showed that breadth developed gradually starting from year 2, with the number of viruses neutralised as well as the antibody titer increasing over time, peaking at 4 years post-infection with little activity thereafter. The extent of cross-neutralising activity correlated with CD4 T cell decline and viral load at 6 months post infection, suggesting that early events set the stage for the development of breadth. Mapping of the epitopes targeted by cross-neutralising antibodies revealed that in most cases they recognised one of the 4 well-defined sites of vulnerability on the HIV envelope. This included the two glycan-dependent epitopes in the V2 and C3 regions of gp120, the CD4 binding site as well as the membrane-proximal external region (MPER). Isolation of monoclonal antibodies from some of these individuals has confirmed the plasma specificities and revealed interesting insights into antibody ontogeny. Furthermore, we have shown how viral escape from earlier strain-specific antibodies contributes to the formation of neutralising antibody epitopes highlighting the dynamic interplay between viral and antibody evolution. Our ongoing work aims to more precisely define the changes in the antibody and viral repertoires that accompany the development of broadly cross-neutralising antibodies with the aim of mimicking these events in a vaccine scenario.