“Big data and algorithms” devices were integrated into the US military in the 2000s. In Afghanistan and Iraq, the armed forces used them to detect “insurgents” within the population and to predict new “insurrections.” These devices have been used to automate detection and prediction, thereby optimizing political decision-making. However, despite the institutional and financial efforts made, the promise to equip military leaders with more efficient engineering than previously was not fulfilled. This failure did not call into question the perpetuation of these programs. In this article, we propose to analyze the integration of “big data and algorithms” devices in the US military sector through this contradiction.