Project Maven (Algorithmic Warfare Cross-Functional Team)

Project Maven is a US Department of Defense program, established in April 2017 as the Algorithmic Warfare Cross-Functional Team, created to apply machine learning and computer vision to military intelligence data, beginning with drone full-motion video in the campaign against the Islamic State (Source: dodcio.defense.gov). Over the following nine years it grew from a data-labeling and object-detection effort into a data-fusion and targeting platform, Palantir's Maven Smart System, which a Deputy Secretary of Defense memorandum of March 9, 2026 directed to become a program of record under the Pentagon's Chief Digital and AI Office (Source: reuters.com). Its history, contractors and operational record are the subject of Katrina Manson's 2026 book Project Maven (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven), and its role in the February 28, 2026 strike on a school in Minab, Iran has drawn congressional questions and an internal Pentagon investigation (Source: bloomberg.com; House Democrats' letter to Secretary Hegseth on civilian harm in Operation Epic Fury (2026)).

Mandate

Deputy Secretary of Defense Robert O. Work's memorandum of April 26, 2017 established the team "to accelerate DoD's integration of big data and machine learning," with the objective of turning "the enormous volume of data available to DoD into actionable intelligence and insights at speed." Its first task was to "augment or automate Processing, Exploitation, and Dissemination (PED)" for tactical unmanned aerial system and mid-altitude full-motion video in support of the Defeat-ISIS campaign. The memo assigned three lines of work: organizing a data-labeling effort and developing or acquiring algorithms; identifying the computing resources needed and a path to field them; and integrating "algorithmic-based technology with Programs of Record in 90-day sprints." The team was to begin with computer-vision algorithms "for object detection, classification, and alerts," then extend to other defense-intelligence mission areas, and to consolidate existing algorithm-based initiatives across the Defense Intelligence Enterprise. The Under Secretary of Defense for Intelligence oversaw it, with the Director for Defense Intelligence (Warfighter Support) as director and an executive steering group including the Joint Staff, the military services and the DoD General Counsel; the team reported directly to the Deputy Secretary (Source: dodcio.defense.gov).

In remarks at a Defense One Tech Summit in 2017, Cukor, then chief of the team, said the Department would field computer-vision algorithms to extract objects from drone imagery by the end of that year, initially covering 38 classes of objects relevant to the fight against the Islamic State. He said "we are in an AI arms race," that "you don't buy AI like you buy ammunition," and that "AI will not be selecting a target [in combat] … any time soon" (Source: war.gov). In a book excerpt, Manson writes that although Maven was presented as a computer-vision project to sort drone footage, Cukor and his backers intended from the outset to use AI to help target people and objects (Source: thewalrus.ca). The Economist's review of her book likewise argues that the "Maven Smart System" label understates a program built for targeting from the start (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven).

History and administration

Early deployments and Google's withdrawal

According to Manson's account as summarized by The Economist, 2017 trials in Somalia produced algorithms that labeled clouds as flying school buses, and 2018 software in Afghanistan identified trees as people and rocks as buildings (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). Google, an early contractor, announced its participation in March 2018; several thousand employees then signed a petition urging chief executive Sundar Pichai to keep the company out of "the business of war," and dozens resigned. On June 1, 2018, Google Cloud chief Diane Greene told employees the company would not seek to renew the contract after it expired in March 2019 (Source: cnbc.com). After Google's exit Palantir became the most important firm in Maven's development (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven); Manson reports that Amazon Web Services, Microsoft, Scale AI and Nvidia were among the other companies drawn in, and that Palantir chief executive Alex Karp later called Cukor the "founding father of AI targeting" (Source: thewalrus.ca). Manson's book characterizes the 2018 protests as having slowed Google's own defense business more than Maven, as competitors absorbed the work (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven).

Per the same account, the program was used in the 2019 operation that killed Islamic State leader Abu Bakr al-Baghdadi in Syria, in the 2020 drone strike that killed Iran's Qassem Suleimani, and in 2021 to track crowd density at the Kabul airfield during the US withdrawal (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). Cukor left the Defense Department in October 2021 for JPMorgan Chase (Source: fedscoop.com). The Marine Corps reprimanded him for overseeing a toxic culture within Maven (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven).

Ukraine

Russia's 2022 invasion of Ukraine moved Maven to industrial-scale use, feeding "points of interest" against a cloud-computing bill of roughly $1 million per month; General Chris Donahue, commander of American land forces in Europe, pioneered its use there. Between late 2022 and 2023 more than 1,500 algorithms were whittled down to about two dozen for Ukraine. Manson reports that accuracy was context-dependent: a model that reached 70% accuracy in Afghanistan dropped to 30% in the Philippines, and the system struggled with Ukrainian snow and with Russian tanks missing their turrets. By 2026 it still produced 10 false detections per square kilometer assessed in Ukraine (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven).

Transfer to NGA and the CDAO

In 2022 the program began a transition in which responsibility for most of its elements was split between the National Geospatial-Intelligence Agency (NGA) and the Pentagon's Chief Digital and AI Office, with certain duties sent to the Office of the Under Secretary of Defense for Intelligence and Security (Source: defensescoop.com). Sources date the handover differently: NGA states that "the GEOINT aspects of Maven were entrusted to NGA in January 2023" (Source: nga.mil), while Breaking Defense reported in May 2023 that the program had been transferred to NGA and the CDAO "last year" and was poised to become an NGA program of record in fiscal 2024. NGA Director Vice Adm. Frank Whitworth said at GEOINT 2023 that under NGA, Maven had "made some of its most significant technological strides and has already contributed to some of our nation's most important operations" (Source: breakingdefense.com). NGA describes its Maven element as integrating computer vision into military analytic workflows "to automatically detect, identify, characterize, extract, and attribute features and objects in imagery and video," and names Rachael Martin as its Maven program lead (Source: nga.mil). NATO's announcement of its own Maven acquisition notes that the Palantir Maven Smart System "should not be confused with the NGA Maven Warfighter Support System" (Source: shape.nato.int).

Program of record under the CDAO (2026)

In a March 9, 2026 letter to senior Pentagon leaders and commanders, Deputy Secretary of Defense Steve Feinberg directed that the Maven Smart System become an official program of record by the close of fiscal 2026 in September, ordered oversight of Maven moved from NGA to the CDAO within 30 days, and assigned future contracting with Palantir to the Army. Feinberg wrote that it is "imperative that we invest now and with focus to deepen the integration of artificial intelligence (AI) across the Joint Force and establish AI-enabled decision-making as the cornerstone of our strategy" (Source: reuters.com). He also wrote that transitioning MSS to a program of record "will provide the stable funding and resourcing necessary for deliberate, sustained development and enterprise-wide integration." Fiscal 2027 budget materials requested more than $1.5 billion to expand access to MSS under a "Joint Force AI-Enabled Headquarters initiative," and Army Col. Molly Solsbury was appointed MSS program director inside the CDAO in mid-2026; a Pentagon official said the department "is continuing to evaluate programmatic options" (Source: defensescoop.com). Whether the program-of-record designation had formally taken effect by the end of September 2026 is not established by the sources reviewed here; DefenseScoop still described it on September 22, 2026 as a transition directed for the end of the fiscal year (Source: defensescoop.com).

The CDAO's chief since January 2026, Cameron Stanley, is described as a Project Maven alumnus (Source: defensescoop.com). In September 2026 he said the Maven Smart System had replaced "six, eight, ten" separate IT systems that military personnel had used to analyze data, and that the Department was extending it to force readiness, supply chain, budgetary considerations, and modeling and simulation (Source: defenseone.com). See Chief Digital and AI Office (CDAO) and Chief Digital and AI Office (CDAO).

Capabilities

According to Manson, Maven fuses photographs, text, radio and electromagnetic-pulse intelligence into target identification, then coordinates response options such as which aircraft and munitions are closest to a target; a NATO official she quotes called it "the Microsoft Windows of warfighting," and the book states that "with a single click by a human, Maven can turn data into ash" (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). An NGA official told her that computer vision raised the number of targets the United States could hit from under 100 a day to 1,000, and that large language models integrated into Maven raised it fivefold to 5,000 a day; on one day in 2022 in Ukraine it found more than 260 potential targets (Source: thewalrus.ca; The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). Manson reports that the Maven Smart System is deployed in every branch of the US military, incorporating more than 150 data feeds and the work of more than fifty companies (Source: thewalrus.ca).

Maven-developed algorithms have also moved onto weapons. Manson names Goalkeeper, a loitering munition, and Whiplash, an explosives-laden jet ski whose early versions were tested in Ukraine via the CIA (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). On data collection, she recounts that Navy P-8 submarine hunters wiped their hard drives after each mission until Maven forced data preservation, and quotes an unnamed defense official: "We're basically watching the PLA all the time… to get AI training data" (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven).

Assessments of the system's capability differ. Cameron Stanley called Maven's capabilities "revolutionary," while Emelia Probasco of Georgetown's Center for Security and Emerging Technology suggested its abilities may be overstated, with much of the advantage coming from data scale and human expertise: "It's not rocket science. I suspect China already has something like it" (Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT)).

Procurement and contractors

The Defense Department signed an initial five-year, $480 million indefinite-delivery/indefinite-quantity contract with Palantir for the Maven Smart System in May 2024, awarded through the Army's Aberdeen Proving Ground and funded by the Office of the Secretary of Defense; Palantir said the work would initially cover Central Command, European Command, Indo-Pacific Command, Northern Command/NORAD and Transportation Command. In May 2025 the Department raised the ceiling by $795 million to nearly $1.3 billion through 2029. A defense official attributed the increase to combatant commands' growing use of MSS "to command and control dynamic operations," and an Army official said "no acquisition decisions have been made" (Source: defensescoop.com). Palantir chief technology officer Shyam Sankar told the House Armed Services Committee in 2024 that Maven had "tens of thousands" of users (Source: reuters.com).

Snapshot: users and contract value

Frontier-model integration and Anthropic

The New York Times reported on April 12, 2026 that "embedded with a military version of Claude, the chatbot made by the A.I. firm Anthropic, Maven helped generate thousands of targets in the opening weeks of the Iran campaign"; the attribution rests on official accounts that Anthropic neither confirmed nor denied (Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT)). Reuters had reported on March 4, 2026, citing two people familiar with the matter, that the Maven Smart System used multiple prompts and workflows built with Anthropic's Claude, and that Palantir would have to replace Claude with another model and rebuild parts of its software after President Trump ordered the government to stop working with Anthropic (Source: reuters.com). Pentagon chief technology officer Emil Michael said in an account published September 11, 2026 that all of Anthropic's technology in the Maven Smart Systems program had been transitioned out months earlier; only the lede of that account was retrievable (Source: insideaipolicy.com). See Anthropic Claude Gov + Pentagon Dispute (2025–2026) and Anthropic v. United States (Pentagon ban challenge).

NATO

On March 25, 2025 the NATO Communications and Information Agency and Palantir finalized the acquisition of Maven Smart System NATO for NATO's Allied Command Operations, announced by SHAPE on April 14, 2025. NATO described it as providing AI applications "from large language models (LLMs) to generative and machine learning" for intelligence fusion and targeting, battlespace awareness and planning; said the procurement took six months from outlining the requirement to acquisition; and expected use to begin within 30 days (Source: shape.nato.int).

Use in the 2026 Iran war

Reuters described Maven in March 2026 as already the US military's primary AI operating system during the strikes on Iran (Source: reuters.com). Adm. Brad Cooper, head of Central Command, attributed the campaign's pace in part to "advanced A.I. tools," and a Pentagon official described the human role in targeting as "left click, right click, left click" (Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT)). Stanley has said Maven enabled the 1,000 targets the United States struck on February 28, 2026 (Source: bloomberg.com), and that it helped the military strike 13,000 targets in 38 days during Operation Epic Fury (Source: defensescoop.com).

Minab school strike

Two Tomahawk missiles struck the Shajarah Tayyebeh Elementary School in Minab, Iran on February 28, 2026, killing more than 150 people including at least 123 children. Officials involved in the Pentagon's internal investigation, whose report has not been released, told Bloomberg that the main military intelligence database still catalogued the site as an Islamic Revolutionary Guard Corps facility; that the site was fed into Maven with that uncorrected classification and emerged as a recommended day-one target; that target-list work that had taken hours in past conflicts was condensed into several minutes through Maven; and that some Central Command personnel relied too much on the AI in Maven, expecting it to flag stale information or inconsistencies. They cited this alongside cuts to civilian-harm mitigation staff and a compressed targeting timeline. A Palantir spokesperson said the company "is not responsible for the underlying data nor identifying intelligence deficiencies" and that there is no evidence its software was at fault; some of Pete Hegseth's advisers have expressed concern that political scrutiny of Maven could disrupt its further adoption (Source: bloomberg.com). The UN Independent International Fact-Finding Mission on Iran said on September 17, 2026 that there were reasonable grounds to conclude the strike amounted to a war crime (Source: news.un.org, as reported in bloomberg.com). The fuller account is at DOD — Department of Defense (AI Deployer).

A March 12, 2026 letter from House Democrats led by Reps. Sara Jacobs, Yassamin Ansari and Jason Crow asked Hegseth what role AI played "in selecting targets, assessing intelligence, and making legal determinations," whether it is subject to human review and at what point, and whether AI "including the use of Maven Smart System" was used to identify the school as a target and whether a human verified it (House Democrats' letter to Secretary Hegseth on civilian harm in Operation Epic Fury (2026)). The Pentagon has not answered such questions publicly, citing the investigation (Source: bloomberg.com).

Post-strike changes

Central Command's changes to its lethal targeting process became public on September 22, 2026. According to US officials and people familiar with the matter, they refine how planned targets are refreshed and vetted, add open-source feeds to track civilian movement, and make dozens of upgrades to the Maven Smart System, including new ways to sort intelligence to help distinguish a building's function. Stanley, whose office oversees Maven, said the upgrades include feeds tracking civilian maritime traffic; Palantir built capabilities into Maven in the days after the strike, including AI agents that re-review underlying intelligence. Some defense officials and experts have warned privately that the chances of another Minab are still too high (Source: bloomberg.com).

Oversight and debate

Program leaders have described Maven as decision support with humans responsible for targeting decisions: Cukor said in 2017 that AI would not be "selecting a target … any time soon" (Source: war.gov), and Palantir says its software does not make lethal decisions and that humans remain responsible for selecting and approving targets (Source: reuters.com). The "left click, right click" description of the human role raises the question, noted in relation to DoD Directive 3000.09, of whether such a role meets the directive's requirements for human judgment (Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT)). Retired Lt. Gen. Jack Shanahan, under whom Maven began, has warned of "a risk of an escalatory spiral where we're in danger of fielding untested, unsafe and unproven systems" (Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT)).

Within the program's own history, Cukor told Manson: "Let's be able to look ourselves in the mirror and make sure we are careful. We have all this tech; are we the best custodians of it?" General Donahue said, "Ultimately, all this stuff will become automated," and the Economist's reviewer asks whether human overseers "can keep up, particularly if something goes wrong" (The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven). Manson writes that Cukor's most controversial decision was pushing the military to use minimally tested systems in active wars, on the argument that deploying AI before it was reliable was the only way to improve it (Source: thewalrus.ca). The wider debate over autonomy in targeting is covered at Autonomous Weapons.

Relationships

• deployed-by: Palantir Technologies — builder of the Maven Smart System; anchor contractor since Google's 2018 exit
• depends-on: DOD — Department of Defense (AI Deployer) — parent department; the Minab account and civilian-harm context
• related: Chief Digital and AI Office (CDAO) / Chief Digital and AI Office (CDAO) — office ordered to oversee Maven from 2026; its chief Cameron Stanley
• related: U.S. Army (AI Deployer) — assigned future Maven contracting under the March 2026 memo
• instance-of: Autonomous Weapons — operational AI targeting in the kill chain
• related: Drew Cukor — founding chief of the Algorithmic Warfare Cross-Functional Team
• related: Katrina Manson — author of Project Maven (2026)
• supports: The AI that transformed American warfare (Economist, May 14 2026) — review of Manson's Project Maven, Mutually Automated Destruction: The Escalating Global A.I. Arms Race (NYT), House Democrats' letter to Secretary Hegseth on civilian harm in Operation Epic Fury (2026)
• related: Anthropic, Anthropic Claude Gov + Pentagon Dispute (2025–2026) — Claude's reported integration in and removal from Maven
• related: DoD Directive 3000.09 — Autonomy in Weapon Systems, Defense / Military — AI Deployment, AI and National Security