What It Is
Project Maven — formally the Algorithmic Warfare Cross-Functional Team (AWCFT), now rebranded as the Maven Smart System (MSS) — is the United States Department of Defense’s primary program for applying artificial intelligence to military operations. Established in April 2017, it was conceived to address a specific bottleneck: the military was collecting far more drone and satellite surveillance footage than human analysts could review. Maven’s original mandate was to deploy computer vision algorithms that could automatically detect, classify, and track objects of military interest — vehicles, personnel, weapons — in that video stream, compressing hours of manual review into seconds of automated flagging.
The system has expanded considerably beyond that original scope. As of 2025–2026, Maven integrates large language models, generative AI tools, and multiple intelligence data streams to support what the Pentagon calls “intelligence fusion and targeting, battlespace awareness and planning, and accelerated decision-making.” It is used across five US combatant commands — CENTCOM, EUCOM, INDOPACOM, NORTHCOM, and TRANSCOM — with approximately 25,000 active military personnel as of early 2026. The program’s administrator has consistently described Maven as a decision-support tool: all strike decisions require human confirmation before execution. Critics have questioned whether that distinction holds under operational pressure.
Why It Matters for AI Governance and Narratives
Project Maven is the clearest existing test case for almost every major contested question in military AI governance: Can AI targeting tools be deployed ethically at scale? Who bears accountability when algorithmic recommendations prove incorrect? How much transparency do commercial AI contractors owe the public when their systems are used in lethal operations? And, most structurally, can democratic oversight institutions keep pace with the rate at which AI is being embedded in military decision-making?
The program also sits at a fault line in the AI narrative contest this observatory tracks. Silicon Valley’s 2018 reckoning over Maven — in which thousands of technology workers publicly refused to build AI for warfare — inaugurated a framing contest between the “responsible AI” commitments of commercial technology firms and the national security imperatives of the defense establishment. That contest is not resolved. In February 2025, Google quietly reversed its 2018 pledge not to build AI weapons systems. In March 2026, Deputy Defense Secretary Steve Feinberg issued a directive designating Maven the “cornerstone” of the Pentagon’s joint command-and-control architecture — a significant escalation in the program’s institutional status. These moves are not separate events; they are chapters in the same narrative about where the commercial AI industry locates itself in relation to the state’s monopoly on force.
Key Facts and Dates
Origins. Maven was established by a memo signed by Deputy Secretary of Defense Robert O. Work on April 26, 2017, with initial funding of approximately $70 million. The program was designed to draw non-traditional AI talent through the Defense Innovation Unit. Its first director was Brigadier General Jack Shanahan.
The Google controversy (2018). Google was among Maven’s first commercial contractors, providing computer vision tools under a contract worth roughly $9 million. In April 2018, approximately 4,000 Google employees signed a petition to CEO Sundar Pichai demanding the company exit the program, stating: “Google should not be in the business of war.” At least a dozen employees resigned. Google declined to renew the contract in June 2018 and published its AI Principles the following week, including a pledge not to develop technologies for weapons use — a pledge it quietly dropped in February 2025.
Palantir and institutional maturation. Following Google’s exit, Palantir Technologies became the primary contractor. The Pentagon formally certified Palantir as sole-source supplier for Maven in May 2023. Maven became a formal Program of Record on November 7, 2023. In May 2024, the DoD signed a $480 million, five-year contract with Palantir; by May 2025, that ceiling had been raised to nearly $1.3 billion through 2029, reflecting what officials described as “growing demand from combatant commands.”
Operational deployment. Maven supported more than 85 airstrikes in Iraq and Syria in February 2024 and has been integrated into US support for Ukrainian forces, where AI is reported to analyze over 50,000 video streams from the front line monthly. The March 2026 Feinberg directive instructs full transfer of Maven’s administration to the Chief Digital and AI Office (CDAO) and positions the system as central to the Pentagon’s broader all-domain command architecture.
Governance gaps. The Brennan Center for Justice has documented that basic information about Maven’s effectiveness, accuracy rates, and legal compliance remains “hidden from Congress and the public.” Georgetown’s Center for Security and Emerging Technology has flagged that the program’s rapid institutional expansion raises oversight questions requiring congressional attention.
Where to Learn More
- DefenseScoop — Feinberg Maven directive (April 2026): https://defensescoop.com/2026/04/03/palantir-maven-feinberg-directive/ — Current reporting on Maven’s elevated status in DoD architecture
- Brennan Center — The Business of Military AI: https://www.brennancenter.org/our-work/research-reports/business-military-ai — Policy analysis of transparency and oversight gaps
- Observer Research Foundation — AI in Real-Time Warfare: https://www.orfonline.org/expert-speak/ai-in-real-time-warfare-lessons-from-project-maven — Operational and governance assessment
- Arms Control Association — Google Renounces AI Work on Weapons (2018): https://www.armscontrol.org/act/2018-07/news/google-renounces-ai-work-weapons — Primary source coverage of the Google controversy