Recent studies demonstrate that veterans exhibit higher suicide risk compared with the general U.S. population. A prior suicide attempt is a well‐documented predictor of suicide death. Despite increased attention to clinical risk factors of suicide and efforts to develop psychosocial interventions to reduce suicide risk, the underlying biological factors that confer this risk are not well understood. This study examined affect‐modulated startle (AMS) during a series of intermixed unpleasant, neutral, and pleasant pictures in a sample of 108 demographically‐matched veterans at low (passive ideators: n = 26) and high risk (active ideators: n = 29; single attempters: n = 28; and multiple attempters: n = 25) for suicide based on the Columbia Suicide Severity Rating Scale. An exploratory aim involved a longitudinal component in a subset of the high‐risk sample that went on to participate in a randomized 6‐month clinical trial. We investigated whether baseline AMS predicts a subsequent suicide attempt at 12‐month follow‐up. Compared with the other three groups, multiple attempters showed greater startle potentiation during unpleasant pictures and deficient overall startle habituation from early to later trials. The groups did not differ in startle during neutral or pleasant pictures, or self‐reported picture valence. Greater startle during unpleasant pictures was associated with greater emotion dysregulation as measured by the Difficulties in Emotion Regulation Scale and a future suicide attempt assessed prospectively at 12‐month follow‐up. These findings suggest that startle potentiation during unpleasant pictures in multiple‐suicide attempters is a promising psychophysiological biomarker of suicide risk and underscore the clinical importance of targeting emotion dysregulation in the treatment of patients at‐risk for suicide.