Using first-principles electronic structure calculations we identify the anion vacancies in II-VI and chalcopyrite Cu-III-VI 2 semiconductors as a class of intrinsic defects that can exhibit metastable behavior. Specifically, we predict persistent electron photoconductivity (n-type PPC) caused by the oxygen vacancy V O in n-ZnO, and persistent hole photoconductivity (p-type PPC) caused by the Se vacancy V Se in p-CuInSe 2 and p-CuGaSe 2 . We find that V Se in the chalcopyrite materials is amphoteric having two "negative-U" like transitions, i.e. a double-donor transition ε(2+/0) close to the valence band and a double-acceptor transition ε(0/2−) closer to the conduction band. We introduce a classification scheme that distinguishes two types of defects (e.g., donors): type-α, which have a defect-localized-state (DLS) in the gap, and type-β, which have a resonant DLS within the host bands (e.g., conduction band). In the latter case, the introduced carriers (e.g., electrons) relax to the band edge where they can occupy a perturbedhost-state (PHS). Type α is non-conducting, whereas type β is conducting. We identify the neutral anion vacancy as type-α and the doubly positively charged vacancy as type-β. We suggest that illumination changes the charge state of the anion vacancy and leads to a crossover between α-and β-type behavior, resulting in metastability and PPC. In CuInSe 2 , the metastable behavior of V Se is carried over to the (V Se -V Cu ) complex, which we identify as the physical origin of PPC observed experimentally. We explain previous puzzling experimental results in ZnO and CuInSe 2 in the light of this model.