Efficient responses to product quality safety incidents and taking effective preventive measures need accurate warning classifications. The warning system model of product quality and safety was constructed from product quality hazards and social risk bearing capability, 12 integrated indices were selected and the significance and calculation method of each index were determined. The weighting factors for each level were determined using an analytic hierarchy process; the classification was determined using a multi-level fuzzy logic method and divided into four grades: red, orange, yellow and blue. Three product quality and safety incidents were used as examples to validate the model. The results show that fuzzy logic method and AHP combined effectively computes the warning classification, and easy to consider various complex factors