While there are >30 antiretroviral (ARV) drugs approved for HIV therapy, there are only limited data on ARVs in pregnancy. The mean lag time from ARV approval to data availability in pregnancy is 5 years; most ARVs receive regulatory approval with only animal data to evaluate potential fetal effects. For low incidence outcomes such as birth defects, data are often only collected post-approval. To determine if a birth defect is associated with a drug or simply reflects the baseline population rate of a defect, the number of required exposures will vary based on the defect population prevalence. To rule-out >2-fold increased risk in overall defects, with 3% population prevalence, 200 early pregnancy exposures are needed, but to rule out >3-fold increased risk in a rare defect like neural tube defects (NTD), with 0.1% population prevalence, 2000 early exposures are needed. Exposure timing is critical, as teratogenic risk is highest very early in pregnancy, before most women recognize they are pregnant, but most reports do not distinguish pre-conception from first-trimester exposure. Post-pregnancy defect reports to pharmacovigilance databases have limitations including reporting bias, case duplication, and lack of denominators. Prospective reports during pregnancy, with follow-up for birth outcome, such as the Antiretroviral Pregnancy Registry, has fewer biases. In 1998, efavirenz (EFV) was approved with a warning on use in pregnancy due to animal data showing central nervous system defects with in utero exposure in primates. Retrospective reports of NTDs in humans increased concern, leading to FDA classification of ‘positive fetal risk’ in 2005; collection of prospective cases over the subsequent 13 years has now shown no increased NTD risk. In contrast, with dolutegravir (DTG), animal data did not raise concerns, but a well-designed prospective active surveillance study in Botswana detected a potential signal of concern for NTD with preconception DTG exposure. In contrast to the delay experienced with EFV, due to active surveillance, significant numbers of already-exposed pregnancies will be collected prospectively over the next 12 months, and with coordinated global efforts to combine additional exposures with denominator data, this signal should be able to be confirmed or refuted within a year. Continuing prospective active birth outcome surveillance is required as new ARVs are introduced into populations including women of childbearing potential.