Giant otters live in highly cooperative groups. Behavioral observations suggest that groups are composed of a dominant reproductive pair and their offspring of previous years. We combined genetic data and long-term ecological information to determine genetic relatedness within and between groups to verify that hypothesis. We genotyped 12 polymorphic loci of 50 otters from 13 groups and two transient individuals. The average relatedness within groups (r = 0.23) was high, but the degree of relatedness varied within the groups, including groups of unrelated individuals, contradicting the current social hypothesis of an exclusively parent-brood model. Negative correlations between kinship and distance between territories were higher in females, and on two occasions, dominant females were replaced by related subordinates of the same group. Solitary transients were males, suggesting a tendency of male-biased dispersal. These data, combined with long-term ecological and behavioral information, indicate that direct benefits, such as alloparental care, and acquisition, inheritance, and defense of high-quality territories may drive the evolution of group living of this endangered social carnivore.