Throughout many East African savannas, the ant-plant Acacia drepanolobium is found in association with several species of symbiotic acacia ants. Although described as a classic example of protective mutualism, our recent investigations into the costs and benefits of this association have shown that ants are ineffective defenders against many types of herbivores, and impose demographic costs that outweigh benefits over a 5-year period. So are these ant symbionts parasites, or are we missing something? In this talk, I’ll outline some recent results that suggest that what we’ve been missing is elephants. These massive herbivores impose catastrophic but rare damage to acacia trees, and our data suggest that ants serve as an expensive “insurance policy” against these agents of strong selection. But the story doesn’t end there, because even mutually beneficial relations may belie underlying conflicts of interest between mutualist partners. I’ll wrap up the talk with some very recent unpublished data suggesting that ants may have the upper hand in this mutualism, achieving colony densities that are higher than optimal for host plants.