# Maven Smart System The **Maven Smart System (MSS)** is the software product that actually fights the war. Where [[Project Maven]] is the Pentagon *program* — the Algorithmic Warfare Cross-Function Team stood up in 2017 — MSS is the deployed *tool suite* that [[Palantir]] built to operationalize it: the interface troops open, the engine that fuses the sensors, and the system that surfaces targets. This page tracks the product; see [[Project Maven]] for the program's history, ethics, and the [[Google Project Maven Walkout]] that handed the work to Palantir. The distinction matters because the two are routinely conflated. Maven began as a 45-person team using computer vision to spot potential targets in "mind-numbing hours of drone surveillance video." MSS is what that experiment became: in the words of Pentagon [[Chief Digital and Artificial Intelligence Office]] chief [[Cameron Stanley]], "the best AI-enabled C2 [command-and-control] capability on the planet." ## What it does Palantir markets MSS as "a live, synchronized view of the battlespace" that "enables synchronized mission planning and execution." Built on the [[Palantir Gotham]] / [[Palantir Foundry]] data layer with [[Palantir AIP]] providing the AI/LLM layer, the system: - **Fuses multi-source intelligence** — video, satellite imagery, signals intelligence, and other technical feeds — into one operational picture. - **Performs automated object detection** and "centralized target identification and management." - **Runs generative-AI tooling** — including "low-code/no-code" builders that let non-programmers assemble their own apps, and semi-autonomous AI "agents" that carry out routine tasks. - **Summarizes intelligence** — sifts vast libraries of reports and drafts summaries or "potential future courses of action." - **Closes kill chains** — connects sensor to shooter; reported kill-chain times as short as ~86 seconds from detection to strike authorization (see [[Kill Chain]]). A separate offshoot of Maven is run by the National Geospatial-Intelligence Agency ([[NGA]]) — "NGA Maven" — distinct from the DoW-wide MSS. ## The LLM layer and the Anthropic paradox MSS is now a generative-AI system measured in *tokens*. Its AI capabilities run partly on large language models — including [[Anthropic]]'s Claude, integrated via [[Palantir AIP]] under the [[Palantir-Anthropic-Amazon Defense AI Deal]]. This produces the defining contradiction of the product: the self-described "safety-first" AI lab has its model embedded in a system built to close kill chains in under 90 seconds. See [[Dual-Use AI]]. That contradiction became an operational problem in 2026, and it has now been resolved **against** Anthropic. The sequence matters, because each step was widely misread as settling the question: - **February 2026** — [[Donald Trump]] and [[Pete Hegseth]] designate Anthropic a supply chain risk after [[Dario Amodei]] refuses to drop the terms-of-service clauses barring mass surveillance of Americans and fully autonomous lethal weapons. The [[Anthropic Government Ban]] bars defense contractors from doing commercial business with Anthropic at all — the clause that reached [[Palantir]] — pointing MSS toward an expensive model swap. - **August 27, 2026** — **Judge Rita F. Lin** (N.D. Cal.) vacates that order, granting Anthropic summary judgment and a permanent injunction, finding the government acted "based on a desire to make a public example out of Anthropic for its 'arrogance' in criticizing the government." **This ruling remains in effect.** - **September 11, 2026** — Under Secretary of Defense for Research and Engineering [[Emil Michael]] says the point is already moot for this product: "All of the Maven Smart Systems or Palantir work has been transitioned months ago." About 90% of the department's classified AI workloads had migrated off Anthropic, with completion targeted for the end of September. He also warns that a *second* designation, under a different statute, is live in the D.C. Circuit and carries no injunction. - **September 25, 2026** — the **D.C. Circuit upholds that second designation, 2–1**. **Gregory G. Katsas** writes for the majority, joined by **Neomi Rao** (both Trump nominees); **Karen LeCraft Henderson** (G.H.W. Bush) dissents. The Department "reasonably feared that Anthropic might manipulate Claude's design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary." The panel says it has "no reason to doubt" Anthropic's "noble intentions" and "no quarrel" with the California ruling, but holds the designation turns "on what Anthropic does, not why Anthropic does it" — a bad motive is not required. Henderson's dissent: the statute does not treat "a contractor's honest and upfront enforcement of restrictions" as a blacklistable supply-chain risk. The ruling lets DoD keep removing Claude and keep barring contractors from using Anthropic products for departmental work; Anthropic is **not** barred from the federal government as a whole. Hegseth on X: "Confirmed: @AnthropicAI = Supply Chain Risk." Anthropic "respectfully disagrees" and is weighing rehearing, en banc review, or the Supreme Court. So the Claude dependency inside MSS is **gone, not preserved** — removed months before either ruling, and now with appellate cover. Two federal courts are split on the same company under two different rules, and the practical question for this page is settled regardless of how that split resolves: MSS runs on other vendors' models. The product's own history is the cleanest evidence for the argument that Palantir's moat is the perimeter and the ontology rather than any particular model inside it — the model layer was swapped under contract pressure and the system kept fighting. See [[Dual-Use AI]]. > [!warning] Corrects an earlier reading > This page previously recorded Lin's August ruling as meaning "the Claude dependency stands, resolved rather than deferred." That was wrong in both directions: the dependency had already been removed, and the litigation was not over. The [[Palantir Watch Log]] entry for 2026-08-28 carried the same error. ## Contract trajectory | Date | Development | |------|-------------| | 2019 | Palantir takes over Maven after Google's exit; builds MSS on Gotham/Foundry | | May 2024 | $480M, five-year MSS contract awarded to Palantir | | April 2025 | NATO acquires MSS for Allied Command Operations — sole-source, no competitive bid | | May 2025 | Pentagon raises contract ceiling to **$1.3B** (from $480M). ⚠️ This row previously also said Maven was "designated a program of record" in May 2025; that claim is unsourced on this page and appears to conflate the ceiling raise with the formal designation, which 2026 reporting places in 2026 (below). Treat May 2025 as the ceiling action only until a primary confirms otherwise | | March 9, 2026 | Deputy Secretary of Defense **Steve Feinberg** signs a memo directing full MSS implementation before the end of FY2026, moving programme oversight from [[NGA]] to the [[Chief Digital and Artificial Intelligence Office]] within 30 days, with the **U.S. Army** taking over future contracting actions with Palantir | | August 2026 | **MSS becomes an official Pentagon program of record** — reported August 25, ahead of the September 30 fiscal-year deadline Feinberg's memo set. Programs of record carry stable multi-year funding and cross-service integration | | FY2027 request | ~**$2.3B over five years** to expand the system | | September 22, 2026 | Pentagon officials put MSS usage at **more than 100,000 personnel**, up from roughly 50,000 in January — growth attributed to **Operation Epic Fury** | ## Combat debut — Operation Epic Fury / Iran The [[Iran War 2026]] (U.S. codename Operation Epic Fury) was the largest real-world deployment of AI-assisted targeting in history — and the first for which the Pentagon released usage metrics. On May 12, 2026, [[Cameron Stanley]] disclosed that Epic Fury "leveraged Palantir's Maven Smart System… to conduct strike missions across the entire battle space, **13,000 targets in 38 days**." During the campaign: - **Unclassified usage** rose 38%; **classified usage** rose 89% month-to-month. - **Peak daily token consumption** rose **4,425%**, hitting roughly **20 billion tokens a day**. - Troops showed an "insatiable appetite" for the tech (see [[Cameron Stanley - insatiable appetite for AI]]). The scale is the story: a targeting tool that in 2017 helped analysts watch drone video was, nine years later, the coordinating brain of a 13,000-target air war. ## The accuracy and accountability problem The candor about capability sits uneasily beside the record on accuracy. As of 2024 testing, Maven's object identification ran at roughly **60% accuracy versus 84% for human analysts** — meaning the AI was *less* accurate at the identification step even as it accelerated the overall kill chain. The first day of the Iran strikes killed at least 175 people at a school adjoining an IRGC base, reportedly due to out-of-date intelligence. Critics note that as targeting accelerates, meaningful human review compresses toward a rubber stamp — and that the Trump administration's erosion of civilian-casualty tracking means the accountability gap is being "programmed into the software." See [[Project Maven]] and [[Autonomous Weapons Debate]]. ## Who benefits - **[[Palantir]]** — prime contractor; MSS anchors the $1.3B+ Maven program and showcases the broader [[Palantir Product Ecosystem]]. - **[[Anthropic]]** — Claude integrated for analysis via [[Palantir AIP]]. - **[[Amazon]]** — AWS GovCloud infrastructure for classified deployment. - **[[Scale AI]]** — data labeling and AI testing feeding the pipeline. - **[[Peter Thiel]]** — Palantir co-founder and [[Founders Fund]] investor across the connected network. ## See Also - [[Project Maven]] — the program and its history - [[Palantir Product Ecosystem]] — where MSS sits in Palantir's product map - [[Palantir AIP]] — the AI/LLM layer - [[Palantir Gotham]] — the underlying data-fusion platform - [[Cameron Stanley]] — the CDAO who disclosed the Iran usage - [[Iran War 2026]] — the combat debut - [[Kill Chain]] — the underlying doctrine - [[Palantir-Anthropic-Amazon Defense AI Deal]] - [[Dual-Use AI]] - [[Defense Tech Ecosystem]]