Battles & Wars › Modern conflicts › AI and data-driven warfare
Project Maven (Pentagon AI targeting)
launched April 2017· Invention or discovery· Technology, codes & intelligence
Well documentedSolid records, physical evidence or scientific consensus.
A Pentagon program begun in April 2017 to use artificial intelligence to analyze drone video and, later, to help pick targets.
See it on the Battles & Wars map →What happened
Deputy Defense Secretary Robert Work set up the Algorithmic Warfare Cross-Functional Team, nicknamed Maven, to help analysts swamped by surveillance video from the war against ISIS. Google worked on it until 2018, when more than 4,000 employees protested and the company dropped out. Palantir later took a leading role. Its Maven Smart System fuses imagery, radar and signals into one map that suggests targets. It helped target US strikes in Iraq, Syria, Yemen and the 2026 Iran war; NATO bought a version in 2025.
Why it matters
It was the Pentagon's first big bet on AI in combat, and its fights over ethics and contractors shaped how Silicon Valley works with the military.
How we know
The April 26, 2017 Pentagon memo, Google employees' open letter, contract announcements such as Palantir's Maven deal, raised to nearly $1.3 billion in 2025, and statements by US commanders.
Still debated
How accurate its target suggestions are, and how much human checking happens under time pressure, is largely classified.
Connections (3)
▸Project Maven (Pentagon AI targeting) analyzes video footage from Drones and precision strikeWell documented
The connection
By 2017 US drones over Iraq and Syria filmed far more video than analysts could watch. Maven's first job was training computer vision to spot vehicles, people and buildings in that footage, so analysts saw the important seconds. It turned drone surveillance from a flood of footage into searchable data.
When and how
The Pentagon, 2017, first against ISIS; Google engineers helped train the image-recognition models until 2018.
How we know
Pentagon officials described this purpose publicly, and Google confirmed its involvement and its withdrawal.
Debates and open questions
How accurate the image recognition proved in combat, and how often analysts overruled it, has not been publicly reported.
▸Project Maven (Pentagon AI targeting) launched the Pentagon's push into AI and data-driven warfareWell documented
The connection
Maven was the Pentagon's first major program to put machine learning into combat operations. It grew from analyzing video to fusing many kinds of intelligence into target suggestions. Its successes and controversies set the pattern for later military AI, including the Pentagon's Chief Digital and AI Office, created in 2022.
When and how
Launched April 2017 under Deputy Defense Secretary Robert Work; later run by the National Geospatial-Intelligence Agency and Pentagon AI offices.
How we know
Official memos, budgets and contract records trace Maven's growth, and US commanders have described its use in operations.
Debates and open questions
Whether Maven speeds decisions without adding errors, and how it weighs the presence of civilians, cannot be judged from public information.
▸Palantir Technologies became a lead contractor for Project Maven (Pentagon AI targeting)Well documented
The connection
When Google quit Maven in 2018 after employee protests, Palantir, which openly welcomed military work, took over a large part. Its Maven Smart System became the program's main platform: a $480 million contract from May 2024 had its ceiling raised to nearly $1.3 billion in 2025. NATO bought a version too.
When and how
Pentagon work from about 2019; NATO acquired Maven Smart System NATO in March 2025.
How we know
Defense Department contract announcements and NATO's acquisition statement; Palantir's early role was reported by journalists before being confirmed.
Debates and open questions
How much of Maven's capability comes from Palantir versus other contractors and government code is not public.