Following the outbreak of COVID-19, scientists rushed to develop vaccines to protect individuals and ferry the world out of the pandemic. Unfortunately, vaccine hesitancy is a major threat to the success of vaccination campaigns. Research on previous pandemics highlighted the centrality of perceived risk and confidence as core determinants of vaccine acceptance. Research on COVID-19 is less conclusive, and frequently it relies on one-country, cross-sectional data, thus making it hard to generalize results across contexts and observe these relationships over time. To bridge these gaps, in this article, we analyzed the association between perceived risk, confidence, and vaccine acceptance cross-sectionally at individual and country levels. Then, we longitudinally explored whether a within-country variation in perceived risk and confidence was correlated with a variation in vaccine acceptance. We used data from a large-scale survey of individuals in 23 countries and 19 time-points between June 2020 and March 2021 and comparative longitudinal multilevel models to estimate the associations at different levels of analysis simultaneously. Results show the existence of cross-sectional relationships at the individual and country levels but no significant associations within countries over time. This article contributes to our understanding of the roles of risk perception and confidence in COVID-19 vaccines' acceptance by underlining that these relationships might differ at diverse levels of analysis. To foster vaccine uptake, it might be important to address individual concerns and persisting contextual characteristics, but increasing levels of perceived risk and confidence might not be a sufficient strategy to increase vaccine acceptance rates.