IC Technology & Surveillance — 2026-04-07

Project Maven Processes 150+ Intelligence Feeds in Iran War as Pentagon Explores AI Provider Replacements

The Pentagon's AI targeting system Project Maven, managed by Palantir, is processing more than 150 simultaneous intelligence feeds and generating 300-500 strike targets daily in the Iran war. The system generated over 1,000 strike options within the first 24 hours of Operation Epic Fury. The Pentagon is actively exploring replacements for Anthropic's Claude model, previously integrated into Maven, from Google, xAI, and OpenAI after Anthropic demanded its model not be used for fully automated strikes or tracking of U.S. citizens. The Iran conflict represents the first large-scale combat test of AI-assisted targeting.

Analysis
The mid-war search for replacements to Anthropic's Claude, which Anthropic pulled over autonomous strike and citizen-tracking red lines, is an unprecedented technology transition under combat conditions. The 300-500 daily target generation rate raises validation questions that go beyond AI capability: at that tempo, human review of each target becomes a bottleneck that pressure to maintain operational speed could erode.
2 sources
  1. AI at war: Five things to know about Project Maven - TechXplore
  2. AI at war: Five things to know about Project Maven - Al-Monitor

View in full brief →

UNCLASSIFIED // OPEN SOURCE