Background This study examined, for the first time, whether attentional biases can be modified in adolescents at risk for depression. Methods The final sample consisted of 41 girls at familial risk for depression, who were randomly assigned to receive six sessions (864 trials) of real or sham attention bias training [Real attentional bias training (ABT) vs. Sham ABT]. Participants who received Real ABT completed a modified dot‐probe task designed to train attention toward positive and away from negative facial expressions; in contrast, girls who received Sham ABT completed the standard dot‐probe task. Attentional biases, self‐reported mood, and psychophysiological responses to stress were measured at pre‐ and post‐training assessments. Results As expected, girls who received Real ABT, but not those who received Sham ABT, exhibited significant increases from pre‐ to post‐training in their attention toward happy faces and away from sad faces. Moreover, adolescents who received Real ABT were buffered against the negative outcomes experienced by adolescents who received Sham ABT. Specifically, only adolescents who received Sham ABT experienced an increase in negative mood and a pre‐ to post‐training increase in heart rate in anticipation of the stressor. Conclusions The current findings provide the first experimental evidence that attentional biases can be modified in youth at risk for depression and further suggest that ABT modulates the heightened response to stress that is otherwise experienced by high‐risk adolescents.