In the last few decades there has been increasing interest in the role of the amygdala in psychiatric disorders and, in particular, in its contribution to the socio-emotional impairments in autism spectrum disorders (ASDs). Given that the amygdala is a component structure of the “social brain,” several theoretical explanations compatible with amygdala dysfunction have been proposed to account for socio-emotional impairments in ASDs, including abnormal eye contact, gaze monitoring, face processing, mental state understanding, and empathy. Nevertheless, many theoretical accounts, based on the Amygdala Theory of Autism, fail to elucidate the complex pattern of impairments observed in this population, which extends beyond the social domain. As posited by the Relevance Detector theory (Sander et al., 2003), the human amygdala is a critical component of a brain circuit involved in the appraisal of self-relevant events that include, but are not restricted to, social stimuli. Here, we propose that the behavioral and social–emotional features of ASDs may be better understood in terms of a disruption in a “Relevance Detector Network” affecting the processing of stimuli that are relevant for the organism’s self-regulating functions. In the present review, we will first summarize the main literature supporting the involvement of the amygdala in socio-emotional disturbances in ASDs. Next, we will present a revised version of the Amygdala Relevance Detector hypothesis and we will show that this theoretical framework can provide a better understanding of the heterogeneity of the impairments and symptomatology of ASDs. Finally, we will discuss some predictions of our model, and suggest new directions in the investigation of the role of the amygdala within the more generally disrupted cortical connectivity framework as a model of neural organization of the autistic brain.