Shifting policies towards legalisation of cannabis for therapeutic and recreational use raise significant ethical issues for health-care providers seeking evidence-based recommendations. We investigated whether heavy cannabis use is associated with persistent harms to the hippocampus, if exposure to cannabidiol offers protection, and whether recovery occurs with abstinence. To do this, we assessed 111 participants: 74 long-term regular cannabis users (with an average of 15.4 years of use) and 37 non-user healthy controls. Cannabis users included subgroups of participants who were either exposed to Δ9-tetrahydrocannabinol (THC) but not to cannabidiol (CBD) or exposed to both, and former users with sustained abstinence. Participants underwent magnetic resonance imaging from which three measures of hippocampal integrity were assessed: (i) volume; (ii) fractional anisotropy; and (iii) N-acetylaspartate (NAA). Three curve-fitting models across the entire sample were tested for each measure to examine whether cannabis-related hippocampal harms are persistent, can be minimised (protected) by exposure to CBD or recovered through long-term abstinence. These analyses supported a protection and recovery model for hippocampal volume (P=0.003) and NAA (P=0.001). Further pairwise analyses showed that cannabis users had smaller hippocampal volumes relative to controls. Users not exposed to CBD had 11% reduced volumes and 15% lower NAA concentrations. Users exposed to CBD and former users did not differ from controls on any measure. Ongoing cannabis use is associated with harms to brain health, underpinned by chronic exposure to THC. However, such harms are minimised by CBD, and can be recovered with extended periods of abstinence.