Project Maven

Project Maven is where the US military’s modern AI-targeting story begins. Formally the Algorithmic Warfare Cross-Functional Team, it was established in April 2017 by Deputy Secretary of Defense Bob Work to apply machine learning and computer vision to the flood of full-motion video coming off surveillance drones. Analysts were drowning in footage; Maven’s task was to train algorithms to detect and classify objects automatically, leaving humans to make the decisions. It was the Pentagon’s first serious, at-scale push to put modern machine learning into an operational intelligence workflow.

Its early years are best remembered for a clash over who should build it. In 2017 and 2018, Google contributed cloud and machine-learning expertise to Maven, but the work triggered an internal revolt: more than 4,000 employees signed a letter objecting to the company’s role in warfare, and in June 2018 Google said it would not renew the contract. The episode became a defining moment in the debate over Silicon Valley’s relationship with the military.

Maven did not disappear. Around 2022 its work split between the National Geospatial-Intelligence Agency and the new Chief Digital and Artificial Intelligence Office , and it became a program of record in late 2023. It now lives on as the Maven Smart System , the productised data-fusion and targeting platform that Palantir builds under a contract whose ceiling reached nearly $1.3 billion in 2025, with more than 20,000 reported users across combatant commands. The system has been linked to strike operations in Iraq, Syria, Yemen, and the Red Sea. From a small object-detection effort, Maven has become the spine of the Pentagon’s targeting enterprise.