Autoantibodies found in cord blood in children who later develop diabetes might be produced by the fetus. If so, continuous autoantibody production would still be expected in these children at one year of age. We decided to determine autoantibodies in cord blood and to see whether they persisted in these children at one year. Autoantibodies against GAD65 (glutamic acid decarboxylase) and IA-2 (tyrosine phosphatase) in cord blood were determined in 2,518 randomly selected children. Forty-nine (1.95%) were positive for GAD65 antibodies, 14 (0.56%) were positive for IA-2 antibodies, and 3 of them were positive for both GAD and IA-2. Four of the mothers of children with GAD65 autoantibodies in cord blood (8.2%) had type 1 diabetes as did 5 mothers of children with IA-2 antibodies (35.7 %), but only 0.4% of the mothers had type 1 diabetes in the autoantibody-negative group (P < 0.001). Information on infections during pregnancy was available in 2,169 pregnancies. In the autoantibody-positive group, 31.5% had an infection during pregnancy, which was more common than in the autoantibody-negative group of 500 children with the lowest values (20.1%; P < 0.04). At one year follow-up nobody of those with positive cord blood had GAD65 or IA-2 autoantibodies. We conclude that most autoantibodies found in cord blood samples of children are probably passively transferred from mother to child. Antibody screening of cord blood cannot be used to predict diabetes in the general population. Infections during pregnancy may initiate an immune process related to diabetes development.