People can use their eyes to direct others’ attention towards objects and features in their environment. A person who consistently looks away from targets is later judged to be less trustworthy than one that consistently looks towards targets, even when the person is a background distractor that viewers are instructed to ignore. This has been used in previous research to experimentally manipulate the trustworthiness of gaze-cueing faces. However, to date, evidence remains unclear as to how the identity of a cueing face may affect how we interpret trustworthiness from gaze cues. Social psychology has long known that we treat people differently on the basis of social groups, and to date, no studies have investigated whether real-world social groups can affect the incidental learning of trust from gaze cues. As in-group members are typically favoured over out-group members, one might expect that invalid in-group faces would be more resilient to devaluation than out-group faces. Alternatively, in-group members are treated as more heterogeneous individuals than are out-group members, and in-group transgressors can sometimes be punished more harshly than out-group transgressors, so one might expect that this would lead to stricter penalties applied to invalid in-group identities. We explore this by using the real-world social category of race, where participants see faces of both their own and another race, half of which provide valid cues and half invalid. We provide the first investigation into whether these real-world differences in the face identities change how participants learn about trust from patterns of gaze behaviour.