We implemented a field experiment, in collaboration with the Comptroller General Office of Chile (CGO) and the NGO Chile Transparente, that assesses whether citizens update beliefs regarding corruption when they are presented with audit information about malfeasance in their municipal government. The treatment delivered on WhatsApp consists of a short information video that reports the results of the CGO 2020 municipal audits; subjects in control receive a placebo video. We measure incentivized pre-treatment corruption beliefs of 5,525 subjects and then post-treatment beliefs one week and one month after the video information treatment. There is a significant treatment effect -- subjects update negatively when they are informed about corruption in their municipality. Updating is only weakly correlated with the amount of malfeasance reported in the subject’s municipality. We find limited evidence that learning is Bayesian -- errors in priors regarding corruption are only correlated with updating for better educated subjects. Updating is not responsive to signals about change in reported municipal corruption levels. Learning in this experiment is not spurious. Treatment effects persist after a one-month period. Subjects are asked to make a voluntary, costly, contribution to their local municipality -- we observe higher donations for treated subjects in municipalities that perform positively in the 2020 audit.