Technology & OSINT — 2026-03-25

Maven AI Targeting System Accuracy of 60% Versus 84% for Human Analysts Raises Questions as Pentagon Mandates Adoption

The Military Times investigation into the Iran school strike revealed previously unreported Maven system performance metrics: approximately 60% overall accuracy versus 84% for human analysts, with accuracy dropping below 30% in adverse conditions. Despite this, Deputy Secretary Steve Feinberg designated Maven as an official program of record mandating adoption across all military branches by September 2026. The Civilian Protection Center of Excellence workforce was cut by approximately 90%, and CENTCOM's civilian casualty assessment team was reduced from 10 to one. Maven generated hundreds of strike coordinates in the first 24 hours of Operation Epic Fury using Palantir's platform with Anthropic's Claude AI for target ranking and automated legal justifications. Over 120 House Democrats questioned whether Maven was used to identify the school that was struck.

Analysis
Prior briefs documented Maven's 8,000+ targeting actions and its designation as a program of record. The 60% vs 84% accuracy gap and the sub-30% adverse-conditions performance are the first concrete metrics reported. Combined with the 90% reduction in civilian protection oversight, the quality assurance gap between AI-generated and human-validated targeting is widening as the mandate expands.
1 sources
  1. Deadly Iran school strike casts shadow over Pentagon's AI targeting push - Military Times

View in full brief →

UNCLASSIFIED // OPEN SOURCE