An important theme in the philosophy of education community in recent years has been the way in which philosophy can be brought to illuminate and evaluate research findings from the landscape of policy and practice. Undoubtedly, some of these practices can be based on spurious evidence, yet have mostly been left unchallenged in both philosophical and educational circles. One of the newer practices creeping into schools is that of "No best friend" policies. In some schools, this is interpreted as suggesting that children should not have just one best friend but a group of good friends. In others, it is interpreted as suggesting that children should forgo having best friends altogether and be friends with everyone. What is common to both is that friendship is seen as somehow "dangerous". This article offers a preliminary examination of what has been referred to as this "dark side" of friendship. Whilst philosophers such as Patricia White have previously alluded to its existence, there has been little philosophical scrutiny in any broad terms elsewhere. I examine three common arguments commonly used to justify "No best friends" practices: that children can be friends with everyone; that young children are developmentally incapable of "real" friendship hence best friendship should be avoided until later age; that only good people can be good friends. I indicate why this unreflective adoption of practices matters so much and why we should be prepared to challenge these cases. I identify practices that we have good evidence to support as making a positive difference in this area.