Abstract The objective of this study was to carry out a quantitative risk assessment (QRA) of Mycobacterium avium subsp. paratuberculosis (MAP) survival in pasteurized milk produced by industrial dairy plants. Data were collected in three dairy plants (A, B and C) located in three different Italian regions and processing 38.75 (plant A), 89.29 (plant B) and 190.56 million litres (Plant C) of milk yearly. Plants A and plant C produce pasteurized milk, soft and hard industrial cheeses and yogurt; plant B produces only pasteurized milk. In-line milk filter (ILMF) samples and/or bulk milk samples were collected from all 569 herds delivering milk to the three dairy plants. Samples were analysed by quantitative real-time PCR (qPCR). The QRA considered the presence of MAP in ILMF and in bulk milk of all the dairy herds delivering milk to the three investigated dairy plants, estimating MAP concentration in raw milk on the basis of these data, the dilution effect due to mixing milk in collecting trucks and in plant silos, and the effect of pasteurization in reducing the MAP load. The expected fraction of litres of pasteurized milk with 0 MAP would be 99.02%, 99.45% and 99.12%, in plants A, B and C respectively, and an overall percentage 0.55% to 0.98% of pasteurized milk having a MAP contamination >0 colony forming units (CFU)/l and 0.04%–0.11% of pasteurized litres with a MAP contamination > 100 CFU/l was predicted. A daily variation was observed in the proportion of MAP-contaminated litres of milk. The study demonstrated that milk in the dairy plants investigated may be a source of MAP exposure for humans. The between-herd and within-herd MAP apparent prevalence in the investigated areas are likely comparable to those in other areas in Italy, Europe and North America, and the results are applicable to other geographical areas.