How does military automation that replaces soldiers with machines affect the frequency and intensity of conflicts? I argue and demonstrate empirically that despite lowering the cost of fighting, military automation does not necessarily make states pull the trigger in every dispute. Rather it has a stronger effect on those conflicts that involve decision-makers' key interests, that is conflicts that would have taken place also in the absence of military automation. The empirical analysis utilizes a novel detailed time-series cross-sectional dataset of various military indicators from over 170 countries over a period of 2000-2020. The quantitative empirical analysis uses diverse measurements of conflict to examine in details the connection between military automation and conflict. The findings suggest that military automation--operationalized as possession of different types of drones--increases involvement in conflicts primarily against opponents that do not possess such capabilities. In addition, there is no increase in new conflicts, but rather escalation of and additional rounds of already existing disputes. These findings shed new light on the debate between `military automation alarmists' and `military automation skeptics'. The former believe that automation induces moral hazard and makes conflicts more likely, whereas the latter dismiss the importance of automation to conflict due to technical limitations and possible ways to neutralize automated systems. The results here suggest that moral hazard is quite limited, but at the same time it affects already existing conflicts in which only one of the belligerents has such capabilities.