Ants show striking diversity in social organization, raising major questions on the proximate and ultimate causes of such variation. The shift from one-queen (= monogyne) societies to multi-queen (= polygyne) societies has long been viewed as a phenotypically plastic response to ecological and social conditions. In contrast to this view, in five independent ant lineages, alternative forms of colony social organization are controlled by supergenes. Supergenes are large groups of linked genes determining compound adaptive phenotypes, like colour morphs, ecotypes, or social forms. In the best-studied socially polymorphic ant species, Solenopsis invicta and Formica selysi, a supergene variant (= haplotype) is exclusively found in multi-queen colonies and does not recombine with the alternative haplotype. How did such supergenes spread? Supergenes might be favoured by natural selection because they link co-adapted alleles that are beneficial in one social form. The absence of recombination ensures that these alleles are transmitted together and prevents maladaptive combinations between alleles. However, supergenes can also spread selfishly, by distorting Mendelian transmission in their favour. Non-recombining regions are indeed prone to harbour selfish genetic elements, which are typically formed by tight linkage of a killer gene, or toxin, and a rescue gene, or antidote. Strikingly, the social supergene haplotypes associated with multi-queen colonies of ants selfishly favour their own transmission by causing gene drive through distinct mechanisms. In S. invicta, the “polygyne” haplotype causes a green-beard effect: Workers that carry this haplotype kill queens that lack it. In F. selysi, the “polygyne” haplotype is a maternal-effect killer: Offspring of mothers that carry this selfish haplotype fail to hatch when they do not inherit a copy of the haplotype. Overall, the ants’ social supergenes induce unexpected combinations of adaptive and detrimental effects across levels of biological organization. On the one hand, they coordinate individual and colony-level traits, such as queen number, mating system, and dispersal. On the other hand, supergenes cause selfish gene drive and lethal effects due to accumulation of deleterious mutations. The evolutionary dynamics of these large groups of linked genes is therefore particularly complex, and explaining their long-term maintenance is challenging. I argue that social supergenes can be drivers of ant evolution because they spread selfishly and show lethal effects. Therefore, an understanding of the unusual properties of the underlying genetic system is needed to explain variation in colony queen number.