We studied the population genetics of two antimicrobial peptide (AMP) loci, called Mytilin B and Mytilus galloprovincialis defensin 2 (MGD2), in the secondary contact mosaic hybrid zone between Mytilus edulis and M. galloprovincialis. The isolation period between the two species was estimated to be approximately 1 million years (range, 0.5 million to 2 million years) long. During this period, coevolution between microbes and the immune system has likely occurred. The secondary contact, which would date back to approximately 25,000 (0-200,000) years, recently allowed these coadaptations to be rearranged through hybridization. Distinctive polymorphisms were uncovered in coding sequences of the two AMP loci such as insertion/deletion of codons or bisubstituted codons. Very low levels of differentiation were observed between populations of the two species at both loci, while other nuclear loci often showed marked structure among the same samples. The absence of population differentiation proved to be the consequence of secondary introgression of highly divergent alleles. While only a few recombinants were observed at the Mytilin B locus, the MGD2 locus showed a high intragenic recombination rate, which increased in the exon coding for the mature peptide. In addition, standard neutrality tests revealed significant deviations from the mutation-drift equilibrium at both loci. These results suggest that either balancing or directional selection is likely to play a role in the evolution of the two AMPs and introgression would be adaptive. However, evidence accumulated at the Mytilin B locus allows neither for identification of the direction of selection nor for any conclusions on whether selection acted directly on the antimicrobial peptide itself. At the MGD2 locus, a spatial variation of polymorphism patterns along the sequence suggests that selection was direct, although the precise nature of the selection (directional vs. balancing) remains unclear. This study concurs with previous reports of an effect of slight selection on AMP genes evolution in other invertebrates, although selection does not necessarily act on the mature peptides.