BACKGROUND: The HIV Prevention Trials Network (HPTN) 083 trial demonstrated that long-acting cabotegravir (CAB-LA) was more effective than tenofovir disoproxil fumarate-emtricitabine (TDF/FTC) in preventing human immunodeficiency virus (HIV) in cisgender men and transgender women who have sex with men. We characterized HIV infections that occurred in the blinded phase of HPTN 083. METHODS: Retrospective testing included HIV testing, viral load testing, quantification of study drugs, and HIV drug resistance testing. RESULTS: Fifty-eight infections were evaluated, including 51 incident infections (12 in CAB arm and 39 in TDF/FTC arm). In many cases (5 in CAB arm and 37 in TDF/FTC arm), infection was associated with low or unquantifiable study drug concentrations. In 4 cases, infection occurred with on-time CAB-LA injections and expected plasma CAB concentrations. CAB exposure was associated with prolonged viral suppression and delayed antibody expression. In some cases, delayed HIV diagnosis resulted in CAB provision to participants with undetected infection, delayed antiretroviral therapy, and emergence of drug resistance; most of these infections would have been detected earlier with viral load testing. CONCLUSIONS: Early detection of HIV infection and prompt antiretroviral therapy initiation could improve clinical outcomes in persons who become infected despite CAB-LA prophylaxis. Further studies are needed to elucidate the correlates of HIV protection in persons receiving CAB-LA.