Source: Bloomberg (paywall)
The U.S. military's dependence on Palantir's Maven targeting system to identify strike locations created a single point of failure that contributed to killing 123 children in February. Military operators treated algorithmic outputs as validated ground truth rather than probabilistic estimates, then organized their entire targeting workflow around that assumption. Defense contractors market AI-assisted decision-making as precision enhancement, but it functioned as justification acceleration—a gap that exposes how enterprise AI systems designed for "efficiency" can amplify rather than mitigate human error at scale.