Research suggests that oxytocin, a neuropeptide implicated in attachment, is a promising clinical tool because it increases affiliation and attachment behaviors, which are reduced in a range of psychiatric disorders. Oxytocin has been recommended as a psychiatric treatment for adolescents, but this remains largely unstudied. Skepticism is warranted, based on mixed findings in adults and absence of data across development. The objective of this study was to examine the effect of intranasal oxytocin on attachment-related and non-attachment-related trust in an interactive game, determining how this effect differs among inpatient adolescents and healthy controls and whether this effect is moderated by attachment security. There were 122 adolescents (ages 12-17; n = 75 inpatient, 70% female, 37% Black, 24% Hispanic, 20% White, and 20% multiracial; n = 46 control, 55% female, 75% Caucasian) randomized to receive self-administered intranasal oxytocin or a placebo and play a trust game with their mother and a stranger over the Internet. Oxytocin only affected the trust game behavior of adolescents when attachment security was moderate or low. At these levels, oxytocin increased the trust of patients, such that their behavior was equivalent to that of healthy controls. Paradoxically, oxytocin reduced the investments of healthy control subjects. This study takes a first step toward determining whether, and for whom, oxytocin may have a trust-enhancing effect and challenges simplistic notions of oxytocin as the attachment-chemical of the brain-pointing instead to differential oxytocin effects based upon clinical status (patient vs. control) and attachment security.