Autonomous vehicles (AV) are envisioned to reduce road fatalities by switching control of safety-critical tasks from humans to machines. Realizing safety benefits on the ground depends on technological advancement as well as the scale and rate of AV adoption, which are influenced by public perceptions. Employing multilevel structural equation modeling, this paper explores differences in perceptions of AV safety across 33,958 individuals in 51 countries. At the individual level, young males report higher perceptions of current AV safety and predict fewer years until AVs are safe enough for them to use. Since young males are more likely to undertake risky driving behavior, their positivity towards AV safety could lead to more rapid manifestations of safety benefits. Urban, fully employed individuals with higher incomes and education levels also report fewer years until AVs are safe to use. The multilevel model identifies country-level effects after controlling for individual characteristics. Developed countries with greater motorization rates and lower road death rates tend to have greater awareness of AVs but are more pessimistic about their present and future safety. Individuals in developing countries that face greater road safety challenges, particularly involving 2- and 3-wheeled vehicles, predict fewer years until AVs will be safe enough for them to use. Higher AV safety perception among the most risk-taking road users and in developing countries coincide with sociodemographic groups and geographic areas facing the greatest road safety challenges and most in need of improvement, highlighting a potential opportunity to reduce the global disparity in road safety.