Humans are strongly dependent upon social resources for allostasis and emotion regulation. This applies especially to early childhood because humans—as an altricial species—have a prolonged period of dependency on support and input from caregivers who typically act as sources of co-regulation. Accordingly, attachment theory proposes that the history and quality of early interactions with primary caregivers shape children's internal working models of attachment. In turn, these attachment models guide behavior, initially with the set goal of maintaining proximity to caregivers but eventually paving the way to more generalized mental representations of self and others. Mounting evidence in non-clinical populations suggests that these mental representations coincide with differential patterns of neural structure, function, and connectivity in a range of brain regions previously associated with emotional and cognitive capacities. What is currently lacking, however, is an evidence-based account of how early adverse attachment-related experiences and/or the emergence of attachment disorganization impact the developing brain. While work on early childhood adversities offers important insights, we propose that how these events become biologically embedded crucially hinges on the context of the child–caregiver attachment relationships in which the events take place. Our selective review distinguishes between direct social neuroscience research on disorganized attachment and indirect maltreatment-related research, converging on aberrant functioning in neurobiological systems subserving aversion, approach, emotion regulation, and mental state processing in the wake of severe attachment disruption. To account for heterogeneity of findings, we propose two distinct neurobiological phenotypes characterized by hyper- and hypo-arousal primarily deriving from the caregiver serving either as a threatening or as an insufficient source of co-regulation, respectively.