Potential interspecific hybrids are usually identified in natural populations by their proximity to interbreeding species or their intermediate phenotypes; hybridization can then be confirmed by comparing the genetic make-up of putative hybrids to pure species. In contrast, detecting interspecific hybridization and misclassifications in ex situ collections can be difficult because fine-scale geographic locations and species-specific phenotypic data are generally unavailable. Thus, there is little a priori information available to suggest which individuals might be hybrids. Instead, hybrids or misclassified individuals must be identified based on molecular data via population assignment and admixture detection programs. We have applied a variety of population assignment and admixture detection programs to over 400 samples of four closely related Malus species held in the USDA-ARS National Plant Germplasm System that were genotyped at 19 SSR loci. Our findings indicate that over 10% of the samples of the wild species M. sieversii and M. orientalis, and nearly 20% of the samples of the wild species M. sylvestris, may be admixed or misclassified. The percentage of admixed or misclassified samples of the domesticated species, M. × domestica, was much lower, at less than 5%. These findings have important implications for how to detect hybridization and misclassification in large collections using molecular data and, ultimately, how to maximize the utility of the collections. In addition, the presence of wild-collected samples that show admixture with domesticated apple suggests that gene flow may be occurring from the crop into natural populations of the wild species.