How AI Is Reshaping the Military: From Drones to Decision Support

01 The Maven Evolution

Project Maven — the AI program that identifies objects in drone footage — started in 2017 with a team of 50 and a budget of $100 million. In 2025, it processes over 50 million images per day across 400 drone platforms with a budget of $2.4 billion. The system can identify 200+ object types (vehicles, buildings, personnel, weapons) in real time. Human analysts now review flagged items instead of raw footage, increasing throughput by 80x.

02 Decision Support, Not Decision Making

The Pentagon's official position is that AI will support, not replace, human decision-making in lethal operations. This is encoded in Directive 3000.09, which requires 'human-in-the-loop' for weapons release. But the line between 'support' and 'decision' is blurring. An AI that recommends a target in 0.3 seconds and a human who approves in 2 seconds is functionally autonomous — the human is a rubber stamp, not a decision-maker.

03 The Swarm Race

Both the US and China are developing autonomous drone swarms — groups of 100+ drones that coordinate without central control. The US's Replicator program aims to deploy thousands of low-cost autonomous drones by 2026. China has tested swarms of 200+ units in exercises. The strategic implication: mass (quantity of units) may overwhelm quality (sophisticated individual platforms). A swarm of 1,000 $10,000 drones may defeat a single $100 million fighter jet.