Combat‐related post‐traumatic stress symptoms (PTSS) are prevalent among recently deployed veterans, making identification of biomarkers of PTSS in this population a public health priority. Given the link between threat processing neurobiology and PTSS, the threat‐related late positive potential (LPP), an ERP reflective of attentional processing sensitive to emotion and its regulation, may have utility as a cost‐effective biomarker. Existing PTSS/threat‐related LPP findings are mixed, possibly due to variability in PTSS across samples, but this has never been explicitly tested. To address this gap, LPP amplitudes to angry, fearful, and happy emotional face stimuli were recorded among 81 combat‐exposed veterans at a VA hospital. A quadratic relationship between self‐reported PTSS and LPP amplitude modulation by angry faces emerged such that greater PTSS was related to a decreased LPP response to angry faces among veterans with subthreshold PTSD and an enhanced LPP response to angry faces among veterans with probable PTSD. These results suggest that prior mixed findings may be due to variability in PTSS severity. In addition, exploratory moderation analysis revealed that PTSS was positively associated with late LPP modulation for veterans reporting low cognitive reappraisal use and negatively associated with late LPP modulation for veterans reporting high cognitive reappraisal use. All results were specific to the 1,000–3,000 ms LPP time window. Thus, the functional nature of LPP modulation by direct threat cues may depend upon PTSS severity and/or related variables (e.g., cognitive reappraisal utilization).