The Palantir Parallel — When Software Becomes an Instrument of State Power

Full report

On this page

Financial Architecture and Economic Engine

The transformation of enterprise software from a back-office administrative tool into a core instrument of statecraft is reflected in the financial structure and public sector contracting footprint of Palantir Technologies Inc. According to Palantir’s SEC Form 10-K for the fiscal year ended December 31, 2025, the enterprise generated \$4.475 billion in total consolidated revenue, representing a 56% expansion from \$2.866 billion recorded in fiscal year 20241. This acceleration was heavily sustained by the domestic market, where overall U.S. revenue expanded by 75% year-over-year to reach \$3.320 billion1. The company’s revenue distribution demonstrates a substantial dependency on sovereign public sector institutions, with the government segment accounting for 54% ($2.416billion)oftotalrevenuein2025,whilethecommercialsegmentgeneratedtheremaining46%($2.058 billion)2. Geographically, 74% of consolidated revenue originated from domestic U.S. customers, with 26% derived from international defense, civil, and commercial operations2. The financial results reported for the fourth quarter of 2025 illustrate an escalating growth trajectory driven by public and private sector demand for automated decision-support systems. Total quarterly revenue reached \$1.407 billion, reflecting a 70% increase year-over-year1. U.S. revenue surged 93% year-over-year in the fourth quarter to $1.076billion,fueledbya137%year-over-yearincreaseinU.S.commercialrevenue($507 million) alongside a 66% year-over-year expansion in U.S. government revenue (\$570 million)1. Guidance issued for fiscal year 2026 projects total consolidated revenue between \$7.182 billion and \$7.198 billion—representing a planned annual growth rate of 61%—with U.S. commercial revenue expected to surpass \$3.144 billion1.

Palantir’s operating model yields cash conversion, reporting an adjusted free cash flow margin of 51% (\$2.270 billion) for fiscal year 2025 and closing the period with \$7.2 billion in cash, cash equivalents, and short-term U.S. Treasury securities without holding long-term debt1. The company achieved a Rule of 40 financial metric score of 127% in the fourth quarter of 2025 by combining a 70% year-over-year revenue expansion rate with a 57% adjusted operating margin1. The core commercial model relies on embedding software platforms deep within client operations to drive retention and expansion, serving 954 total enterprise customers by year-end 2025, with its top twenty customers generating an average trailing-twelve-month revenue of \$93.9 million2. This unit economic performance is driven by a forward-deployed engineering deployment paradigm6. Rather than distributing off-the-shelf software packages through traditional channel partners, Palantir embeds software engineers directly within client operational commands, military units, intelligence directorates, and civil administrative agencies6. These forward-deployed engineers clean data feeds, construct semantic ontologies, and program specialized operational workflows directly inside the client's infrastructure6. This deployment strategy moves institutions through three distinct phases: Acquire (pilot deployments and ontology construction), Expand (operational integration across core business units), and Scale (enterprise-wide reliance where the platform functions as the institution's primary operational operating system)6. Once institutionalized, the software ontology forms the primary conduit through which administrative and command decisions flow, isolating the vendor from standard software replacement cycles8.

Product Architecture and Core Platforms

Palantir’s platform architecture is structured across four primary systems: Gotham, Foundry, the Artificial Intelligence Platform (AIP), and specialized tactical software suites including the Maven Smart System6. Palantir Gotham was built specifically for defense, intelligence, and law enforcement environments11. Gotham functions as an intelligence synthesis and entity-resolution system capable of ingesting vast volumes of unstructured and structured data across isolated government repositories11. The platform aggregates signals intelligence, satellite imagery, field report logs, biometrics, financial transactions, and intercept records into a unified link-node graph11. By resolving disparate records into singular real-world entities—such as individuals, vehicles, locations, organizations, and events—Gotham provides defense analysts and military commanders with a consolidated common operating picture10. Palantir Foundry operates as an enterprise-wide data operating system across commercial, industrial, and civil public sector institutions6. At the foundation of Foundry lies the Ontology, a semantic abstraction layer that translates raw, disconnected enterprise datasets—drawn from relational SQL databases, legacy enterprise resource planning platforms, cloud data lakes, and edge sensor feeds—into digital representations of physical operations6. Within the Ontology, institutional assets, operational processes, and business logic are modeled as Objects, Properties, and Links9. A hospital network’s ontology represents beds, physicians, clinical interventions, and patient records as interconnected objects; similarly, a civil aviation authority models aircraft, maintenance logs, flight paths, and air traffic sectors9. Foundry supports zero-copy data integration, dynamic pipeline orchestration, and continuous state-synchronization, allowing operational teams to simulate interventions and execute administrative actions directly from within the user interface6. The Artificial Intelligence Platform binds large language models and generative artificial intelligence tools directly to the structured logic of the Foundry and Gotham Ontologies6. Unconstrained general-purpose language models frequently suffer from hallucinations, lack operational context, and introduce data-leakage vulnerabilities6. AIP mitigates these failure modes by binding artificial intelligence agents to the security primitives, access control policies, and action frameworks defined within the Ontology6. The technical architecture of AIP consists of four core operational components:

• AIP Logic: An environment that allows developers and operational analysts to build, test, and deploy prompt chains and agentic tools that query the Ontology to execute multi-step administrative processes6.
• Agentic Runtime and Action Framework: An execution layer that enables software agents to propose or automatically execute write-back actions into connected enterprise systems—such as updating legal case management files, re-routing logistics allocations, or generating military target nominations—subject to predefined administrative guardrails and mandatory human-in-the-loop checkpoints6.
• Security Primitives and Governance: Inherits granular access control lists, role-based permissions, classification markings, and purpose-based access controls directly from the underlying data layer6. If an individual operator or automated software agent lacks formal clearance for a specific object property, that property is dynamically masked during model context retrieval6.
• AIP Evals: A continuous testing and evaluation framework designed to benchmark non-deterministic model outputs against empirical unit tests, detect performance drift, and maintain immutable audit logs for regulatory compliance9.

The Maven Smart System represents the operationalized evolution of the Department of Defense’s Project Maven, established by the Defense Innovation Unit in 2017 to process tactical drone video feeds using computer vision algorithms13. Following public ethical pushback and non-renewal by Google, Palantir assumed primary software development for the program19. The Maven Smart System aggregates data feeds from satellite constellations, open-source social media communications, signals intelligence, and radar telemetry into an AI-enabled command-and-control platform10. In May 2024, the U.S. Army awarded Palantir a \$480 million contract expansion to scale Maven Smart System tools across global combatant commands, bringing cumulative Maven-related contract obligations to \$1.3 billion8. The system was formally designated as an official Defense Program of Record by Deputy Secretary of Defense Steve Feinberg, securing long-term defense funding and mandating its operational integration across all military branches by late 202618. Maven serves as the software core for Combined Joint All-Domain Command and Control, automating threat detection, battle damage assessment, and strike resource allocation18.

Documented Capability Matrix

The technical capabilities, primary operational domains, software mechanisms, and governance frameworks across Palantir’s platform portfolio are detailed in the matrix below:

Government Contracting Footprint and Public Sector Penetration

Palantir’s expansion across federal defense, intelligence, and civilian administrative institutions is sustained by billions of dollars in obligated public sector contracts8. Within the Department of Defense, the U.S. Army awarded Palantir a major Enterprise Agreement contract consolidation on July 31, 20258. This agreement combined 75 separate procurement vehicles into a unified software contract with a funding ceiling of up to \$10 billion, embedding Palantir’s platforms across all operational army echelons8. Additional major defense programs include Army Vantage—a \$618.9 million data platform designed to synthesize personnel readiness, equipment logistics, and financial accounts into a real-time executive dashboard—and the Tactical Intelligence Targeting Access Node (TITAN) program8. Under its \$178.4 million TITAN award, Palantir manufactures mobile, AI-enabled ground stations that process sensor feeds downlinked directly from satellite constellations, unmanned aerial systems, and terrestrial sensors to deliver target acquisition data directly to field artillery units8. Across federal civil and homeland security agencies, Palantir has constructed a extensive administrative data architecture8. The company has provided data analytical infrastructure to U.S. Immigration and Customs Enforcement since 2011, securing over \$287 million in cumulative contract obligations8. Early systems included Homeland Security Investigations' Investigative Case Management (ICM) system—supported by a sole-source contract that expanded beyond \$145 million—and the FALCON mobile application suite utilized for workplace enforcement operations and gang investigations23. In April 2025, ICE awarded Palantir a \$30 million sole-source contract for ImmigrationOS (the Immigration Lifecycle Operating System)8. ICE contract documentation specifies that ImmigrationOS was designed to streamline target identification, apprehension tracking, and removal enforcement, providing real-time operational visibility into both forced deportations and instances of self-deportation23. Concurrently, ICE operationalized ELITE (Enhanced Leads Identification & Targeting for Enforcement), an automated targeting system that ingests non-traditional administrative datasets—including state Medicaid enrollment files, passport application databases, and commercial location records—to generate probabilistic target dossiers assigned with algorithmic confidence scores8. Palantir’s civil agency expansion extends across financial and public health infrastructure8. Under reported \$1.5 billion contracting initiatives linked to the Internal Revenue Service and the Department of Government Efficiency (DOGE), Palantir was selected to construct a unified API layer ("Mega API") across historically isolated tax databases, asset disclosure records, and compliance filings8. Within healthcare, the company holds a \$443 million enterprise platform contract with the Centers for Disease Control and Prevention to maintain national outbreak surveillance and medical countermeasure logistics8. In late 2025, contracts awarded by U.S. Citizenship and Immigration Services expanded Palantir’s software footprint into the legal benefits and naturalization processing mechanisms of the federal government27. Palantir's public sector strategy has also secured critical overseas government contracts29. In November 2023, NHS England awarded Palantir a seven-year, £330 million (\$441 million) contract to build and operate the Federated Data Platform (FDP)29. The FDP unifies patient health records, surgical waiting lists, and emergency department telemetry across regional healthcare trusts29. Concurrently, the company holds contracts valued at over £240 million with the UK Ministry of Defence29. However, this international expansion has encountered public resistance29. In 2025, London Mayor Sadiq Khan intervened to block a proposed £50 million contract that would have deployed Palantir software across the Metropolitan Police Service to process investigative criminal intelligence29. Palantir initiated High Court legal proceedings against the Mayor’s Office for Policing and Crime, alleging unfair exclusion, before both parties agreed to withdraw claims in October 2026 with the contract remaining blocked29.

Vendor Claims versus Independent Scrutiny and Civil-Liberties Analysis

Palantir markets its software platforms as objective, neutral administrative foundations designed to enhance organizational efficiency, enforce strict data privacy through cryptographic access controls, and preserve human oversight over complex operations6. Conversely, civil liberties organizations, investigative journalists, and independent research bodies present an opposing analysis, asserting that these platforms create pervasive surveillance networks, encourage procedural automation bias, and weaken administrative transparency8.

Civil liberties advocacy groups, including the Electronic Frontier Foundation (EFF) and the American Civil Liberties Union (ACLU), argue that Palantir’s primary software innovation is not purely algorithmic, but rather its capacity to dissolve legal and technical data boundaries8. Historically, statutory frameworks such as the Privacy Act of 1974 placed strict boundaries on federal agencies, prohibiting them from merging isolated administrative databases to prevent the creation of domestic surveillance profiles on citizens24. Palantir’s semantic ontology frameworks bypass these historical barriers by linking state driver's license records, tax filings, Medicaid enrollment rolls, passport files, and commercial data broker feeds into a single queryable graph8. In the context of immigration enforcement, lawsuits brought by 404 Media and the Freedom of the Press Foundation uncovered that the ELITE platform ingested state Medicaid health data to locate deportation targets8. The EFF criticized the integration of public health records into deportation systems as a breach of taxpayer trust and data privacy rights8. Furthermore, internal training manuals obtained by The Guardian confirmed that the FALCON mobile application enabled field agents to perform real-time GPS tracking and organizational network mapping during enforcement operations28. In the criminal justice arena, independent investigative reporting revealed that Palantir operated an unmonitored predictive policing program with the New Orleans Police Department for six years8. The software used intelligence data to identify individuals statistically likely to commit or become victims of violent crime8. Defense attorneys were never informed that Palantir software had generated the target dossiers or investigative leads used against their clients8. Legal scholars note that withholding the involvement of predictive software deprives criminal defendants of their right to challenge algorithmic bias, input data corruption, or source validity, directly violating constitutional discovery mandates established under Brady v. Maryland8. In the United Kingdom, the deployment of Palantir software across the National Health Service triggered public controversy29. Privacy watchdogs, including Medact and Foxglove, revealed that NHS England signed unredacted FDP contracts prior to finalizing legally binding data protection safeguards29. NHS England subsequently acknowledged that Palantir engineers had accessed identifiable patient records, prompting over 44,000 citizens to submit formal data protection objections29. An empirical study conducted by Foxglove revealed that one-third of NHS hospital trusts using Palantir tools recorded a net reduction in surgical operations after platform implementation, contradicting official claims regarding software-driven efficiency gains29.

Software Logic versus Executive Responsibility

The embedding of agentic decision-support software across military, intelligence, and civil operations reshapes the relationship between elected officials, civil servants, and private technology providers18. As algorithms assume responsibility for synthesizing operational data, assessing national security threats, and recommending administrative interventions, the distinction between software recommendations and constitutional state authority becomes blurred8. Automation bias represents the primary cognitive mechanism through which software logic exercises control over human decision-makers8. In high-stress operational environments—such as air defense command centers, law enforcement fusion centers, or social service triage boards—operators face intense cognitive demands8. When software interfaces present probability scores, link charts, and target selections labeled as "high confidence," human operators exhibit a psychological tendency to accept the algorithm's output without independent verification8. Rejection of a software-generated recommendation requires an operator to actively override a system backed by massive data synthesis8. If an operator overrides a high-confidence recommendation and an operational failure occurs, the human operator bears sole personal responsibility8. Consequently, statutory "human-in-the-loop" safeguards frequently degrade into procedural formality, transforming analytical recommendations into de facto administrative decisions8. To preserve democratic governance, analytical software recommendations must be distinguished from legal authority, military targeting, and executive responsibility:

• Analytical Recommendations: Probabilistic software outputs, risk classifications, entity link graphs, and automated resource optimizations generated by data synthesis platforms6. These recommendations represent mathematical inferences derived from input data; they possess no statutory standing or constitutional power24.
• Legal Authority: Vested statutory power granted exclusively to sworn public officials, judicial officers, and civil servants by constitutional or legislative mandate24. Legal authority empowers officers to deprive individuals of liberty, seize assets, collect taxes, or issue administrative orders under strict procedural due process24. Private technology providers and software algorithms cannot possess or exercise legal authority24.
• Military Targeting Decisions: Sovereign decisions to apply lethal combat force against designated target nominations, governed strictly by the Law of Armed Conflict, international humanitarian law, and established operational Rules of Engagement18. The authority to execute strike operations remains restricted to authorized human commanders within the legal chain of command18.
• Executive Responsibility: Non-delegable constitutional accountability borne by elected executive leaders and appointed department heads24. Executive leaders remain politically and legally responsible for all actions executed by state agencies under their direction, regardless of whether those actions were executed via human agency or automated software systems24.

When state institutions rely on enterprise software to formulate policy, execute law enforcement actions, or nominate military targets, an accountability vacuum emerges18. Public officials routinely defend controversial decisions by asserting that their actions were guided by "objective, data-driven software models"23. Conversely, technology vendors disclaim legal responsibility for real-world harms by asserting that their platforms function merely as neutral decision-support tools that leave ultimate operational discretion to government operators23. This dynamic splits administrative accountability, allowing public officials and technology vendors to shift responsibility while citizens suffer the consequences of automated state power8.

Application: Eviulon Services and Contractor Evulgare

To examine these institutional dynamics, consider a comparative scenario featuring Eviulon, a fictional machine-intelligence software vendor, and Evulgare, an independent forward-deployed engineering contractor. Eviulon develops core machine-intelligence platforms, offering multi-modal foundational models, dynamic ontology engines, and automated workflow orchestrators6. Eviulon licenses its base software stack to government institutions, maintaining a corporate posture that its algorithms are domain-agnostic, configurable, and built in compliance with standard administrative governance frameworks9. Evulgare functions as an embedded forward-deployed engineering integration firm. Evulgare’s software engineers embed directly within state agencies, military command centers, domestic intelligence bureaus, and tax authorities6. Operating inside classified or secured perimeters, Evulgare engineers clean agency data streams, construct custom operational ontologies, and program specialized agentic tools that operationalize Eviulon’s software stack for real-world government execution6. This operational division reveals three structural governance failures:

1. Contractual Shielding of Sovereign Policy Choices: When the state licenses Eviulon’s base software and tasks Evulgare with building an automated enforcement pipeline, the state effectively outsources the structural logic of its administrative authority8. If Evulgare configures an algorithm that prioritizes low-income citizens for tax audits or welfare revocations based on geographic and relational indicators, public officials attribute the targeting criteria to "objective algorithmic risk models"8. The state shields its policy choices behind Evulgare’s proprietary configuration scripts8.

2. The Erasure of Judicial Auditability: Citizens seeking judicial redress against automated enforcement actions encounter an impenetrable legal barrier8. Courts attempting to review the state's decision-making process face corporate trade-secret protections asserted by Eviulon alongside national security classification claims asserted by Evulgare8. The citizen is denied access to the underlying code, model weights, and data linkages that generated the adverse administrative decision, rendering procedural due process ineffective8.

3. De Facto Executive Authority Vested in Private Engineers: Evulgare engineers occupy a pivotal position between raw public data and sworn civil servants6. By deciding how entities, attributes, and relationships are structured within Eviulon’s ontology, Evulgare engineers dictate what data public officials see, how options are framed, and which targets are highlighted for enforcement9. Despite holding no statutory commissions or democratic mandates, Evulgare’s staff exercise de facto executive authority over the operations of the state24.

Competing Assessments of Benefits and Risks

The deployment of integrated enterprise software across public institutions introduces sharp trade-offs between administrative efficiency and constitutional governance.

From an institutional perspective, integrated enterprise software solves persistent data fragmentation within large government bureaucracies6. Modern state functions—ranging from military operations across dispersed theaters to public health responses during pandemics—require the rapid synthesis of vast data streams10. By unifying disconnected legacy databases into a semantic ontology, platforms like Foundry and Gotham enable public administrators to query complex operational systems in real time, eliminate administrative redundancies, and allocate public resources with high precision6. Conversely, from a civil liberties and constitutional governance perspective, the deep integration of these software platforms creates systemic risks8. Data fragmentation historically served as a deliberate constitutional buffer designed to prevent the concentration of state power24. Dissolving these technical boundaries through commercial data platforms enables automated surveillance structures that operate without individualized judicial warrants8. Furthermore, as state agencies become dependent on proprietary software ecosystems for daily operations, public institutions surrender core technical competence to commercial entities8. This dependency allows technology vendors and embedded contractors to exercise structural influence over state policy, compromising the independence of public governance8.

Fictional Revelations: Citizen Discoveries of Software-Driven Governance

The six narrative scenarios below illustrate how citizens uncover consequential state decisions driven by integrated decision-support platforms, automated risk engines, and embedded contractor operations.

Revelation 1: The Automated Municipal Resource Grid Revocation

An investigative exposure revealed that a major metropolitan utility board deployed an integrated enterprise software platform to manage municipal water and electrical grid distributions. While publicly marketed as an environmental optimization initiative to prevent power surges, internal documents disclosed that embedded contractor engineers had configured an automated risk scoring engine. The software continuously ingested municipal utility payment records, state driver's license updates, public assistance records, and commercial mobile phone location feeds captured near peaceful political demonstrations. Citizens who participated in protests opposing local tax increases were assigned elevated "infrastructure strain risk scores." The software automatically triggered utility disconnections for high-risk households under the pretext of emergency grid conservation. Disconnections occurred without human administrative review or advance notice, leaving thousands of residents without electrical power or clean water until independent journalists audited the platform's configuration scripts.

Revelation 2: The Probabilistic Child Protective Service Removal Engine

A child welfare advocacy group uncovered that a state family services department had operationalized an automated AI risk assessment platform designed to predict child neglect. The software did not merely advise caseworkers; it automatically generated emergency child removal petitions submitted directly to family court judges. The platform ingested sensitive Medicaid diagnostic records, prescription drug monitoring histories, food assistance transaction logs, and unverified anonymous tip line recordings. The software's machine-learning model applied disproportionate statistical weight to low-income indicators, automatically generating emergency removal orders for families suffering from temporary financial distress. Caseworkers admitted during depositions that they routinely signed and filed the software-generated petitions without reviewing raw evidence, relying entirely on the platform's high-risk classification score to execute child removals, violating fundamental parental due process rights.

Revelation 3: The Secret Airspace Transit Watchlist Expansion

A commercial airline passenger holding dual citizenship was systematically barred from boarding domestic flights over a twelve-month period without administrative explanation. A civil liberties lawsuit forced the government to disclose the existence of an unmonitored integration layer operating between federal aviation security databases, private credit reporting bureaus, and a commercial intelligence software platform. The platform had automatically classified the citizen as a national security threat after linking a wire transfer sent to overseas family members for medical care, the installation of an encrypted messaging application, and metadata from a hotel booking in a foreign capital. The platform added the individual to a zero-board security watchlist without human judicial review or agency oversight. During court proceedings, government attorneys acknowledged that agency officials could not explain the mathematical weights used to generate the security flag due to proprietary trade-secret protections asserted by the software vendor.

Revelation 4: The Automated Tax Audit and Benefit Asset Freeze Matrix

Thousands of small business owners received simultaneous legal notices confirming that their commercial bank accounts had been frozen and their tax-exempt status suspended by state revenue authorities. An audit conducted by a state comptroller revealed that the tax authority had deployed an automated risk matrix managed by an embedded defense contractor. The platform continuously processed point-of-sale credit card transactions, supply chain customs manifests, and social media relationship graphs. The software calculated a daily "tax divergence score" for registered businesses. If a business score breached an algorithmically determined threshold, the platform automatically issued administrative bank garnishments without requiring a human tax auditor to inspect the business's accounting books. Business owners were forced to shoulder heavy legal expenses to prove their compliance and unfreeze their operating accounts, driving dozens of innocent enterprises into bankruptcy.

Revelation 5: Unannounced Border Drone Strike Targeting Matrix

Whistleblower disclosures revealed that tactical strike operations conducted along an unannounced foreign border zone were driven by an agentic military command platform. The software integrated satellite imagery streams, signals intelligence intercepts, and open-source social media communications to generate automated target nominations for unmanned aerial vehicle operators. The system assigned a numerical confidence percentage to suspected hostile combatants. Under modified operational rules of engagement approved by military leadership, any target package exceeding an 85% algorithmic confidence score received automatic strike authorization. The whistleblower confirmed that a software update introduced a pattern-recognition defect that misidentified civilian transport trucks as military supply vehicles based on dust plume telemetry. Human drone operators, working under high operational pressure, approved lethal strikes within seconds of system notification without verifying raw intelligence feeds, leading to civilian casualties along the border corridor.

Revelation 6: Algorithmic Watchlist Categorization of Environmental Activists

A freedom of information request uncovered internal communications between a state homeland security fusion center and an embedded software contractor team. The documents revealed that an enterprise graph analytics platform had categorized peaceful environmental advocacy organizations as "domestic threat networks." The software automatically constructed entity graphs linking non-profit donor databases, public protest permit applications, and automated license-plate reader logs captured near industrial facilities. When citizens attended multiple peaceful rallies, the software linked their profiles to extremist groups, automatically elevating their threat classification within law enforcement networks. This automated categorization resulted in peaceful citizens being subjected to financial monitoring, airport travel delays, and covert intelligence tracking without the authorization of a judicial warrant or statutory agency directive.

Sovereign Discretion in the Age of Enterprise Software

The transformation of software into an instrument of state power introduces a structural shift in public administration and national security governance. As demonstrated by Palantir’s fiscal year 2025 financial performance—generating \$4.475 billion in revenue, holding \$7.2 billion in liquid assets, and securing multi-billion-dollar defense agreements—the modern state has become deeply reliant on private software ecosystems1. Platforms like Gotham, Foundry, AIP, and the Maven Smart System provide public institutions with unprecedented capabilities to integrate data, monitor operations, and execute rapid interventions6. However, these technological capabilities alter the traditional foundations of democratic accountability. By dissolving statutory data boundaries, automating administrative decision pipelines, and fostering procedural automation bias among public officials, these platforms challenge the exercise of sovereign discretion8. As operational control shifts toward proprietary algorithms and embedded contractor teams, the traditional safeguards of administrative due process, judicial review, and executive accountability face systemic erosion8. Preserving constitutional governance in this era requires establishing strict boundaries between automated analytical recommendations and non-delegable legal authority, ensuring that the critical decisions of the state remain subject to transparent human judgment and democratic control18.

Works cited

1. sec.gov \- a2025q4ex991earningsrelease.htm, https\://www\.sec.gov/Archives/edgar/data/1321655/000132165526000004/a2025q4ex991earningsrelease.htm

2. 2025 FY PLTR 10-K \- Palantir Investor Relations, https\://investors.palantir.com/files/2025%20FY%20PLTR%2010-K.pdf

3. Fourth Quarter of Fiscal Year 2025 \- Palantir IR \- News, https\://investors.palantir.com/news-details/2026/Palantir-Reports-Q4-2025-U-S--Comm-Revenue-Growth-of-137-YY-and-Revenue-Growth-of-70-YY-Issues-FY-2026-Revenue-Guidance-of-61-YY-and-U-S--Comm-Revenue-Guidance-of-115-YY-Crushing-Consensus-Expectations/

4. Palantir Technologies Inc. SEC 10-K Report — Berita TradingView, https\://id.tradingview.com/news/tradingview:962b0ed71cbfc:0-palantir-technologies-inc-sec-10-k-report/

5. Q4 2025 \| Letter to Shareholders \- Palantir, https\://www\.palantir.com/q4-2025-letter/en/

6. What Is Palantir AIP? How It Works, Features, Use Cases Guide, https\://www\.instinctools.com/blog/palantir-aip/

7. Palantir AIP Consulting & Implementation \- BD Emerson, https\://www\.bdemerson.com/service/palantir-aip-forward-deployed-engineering

8. Palantir Government Contracts: \$5.2B Obligated, by Agency, https\://fed-spend.com/blog/palantir-government-contracts-deep-dive

9. Platform overview \- Palantir, https\://palantir.com/docs/foundry/platform-overview/overview/

10. Maven Smart System: Military AI and NATO C2 \| SynapseCommand®, https\://synapsecommand.com/news/maven-smart-system-military-ai-c2

11. (PDF) Mapping Interfacial Regimes of Control: A Qualitative Analysis, https\://www\.researchgate.net/publication/342263230\_Mapping\_Interfacial\_Regimes\_of\_Control\_A\_Qualitative\_Analysis\_of\_America's\_Post-911\_Security\_Technology\_Infrastructure

12. how U.S. immigration enforcement uses data and AI to find and, https\://gist.github.com/ruvnet/da9609939270e3870de465a63352b005

13. New contract expands Maven AI's users 'from hundreds to, https\://breakingdefense.com/2024/05/new-contract-expands-maven-ais-users-from-hundreds-to-thousands-worldwide-palantir-says/

14. Connecting AI to Decisions with the Palantir Ontology, https\://blog.palantir.com/connecting-ai-to-decisions-with-the-palantir-ontology-c73f7b0a1a72

15. Platform Governance on Palantir: A Guide to Secure Enterprise AI, https\://ethicrithm.com/platform-governance-on-palantir-a-guide-to-secure-enterprise-ai-scaling/

16. Securing Agents in Production (Agentic Runtime, \#1) \- Palantir Blog, https\://blog.palantir.com/securing-agents-in-production-agentic-runtime-1-5191a0715240

17. AI Systems Governance through the Palantir Platform, https\://blog.palantir.com/ai-systems-governance-through-the-palantir-platform-74e52d95bdb5

18. Pentagon Officially Designates Palantir's Maven AI as Key Military, https\://mlq.ai/news/pentagon-officially-designates-palantirs-maven-ai-as-key-military-system/

19. Project Maven \- Wikipedia, https\://en.wikipedia.org/wiki/Project\_Maven

20. Palantir wins contract to expand access to Project Maven AI tools, https\://www\.defensenews.com/artificial-intelligence/2024/05/30/palantir-wins-contract-to-expand-access-to-project-maven-ai-tools/

21. Palantir Expands Maven Smart System AI/ML Capabilities, https\://investors.palantir.com/news-details/2024/Palantir-Expands-Maven-Smart-System-AIML-Capabilities-to-Military-Services/

22. Palantir GxP Solutions, https\://www\.palantir.com/offerings/gxp-solutions/

23. Does Palantir Work With ICE?, https\://askpalantir.com/pages/does-palantir-work-with-ice.html

24. All the Ways Palantir is Assisting Trump's Abusive Removal Campaign, https\://www\.aclu.org/news/privacy-technology/palantir-deportation-roundup

25. human rights risks briefing: palantir technologies, https\://investorsforhumanrights.org/sites/default/files/attachments/2019-11/FINAL%20Investor%20Alliance\_Human%20Rights%20Risk%20Briefing%20-%20Palantir.pdf

26. Palantir \- Q3 2025 Investor Presentation, https\://investors.palantir.com/files/Palantir%20-%20Q3%202025%20Investor%20Presentation.pdf

27. Palantir Gotham: From 9/11 to AI \- Medium, https\://medium.com/@KJALaw/palantir-gotham-from-9-11-to-ai-d875d039d55b

28. Documents offer rare insight on Ice's close relationship with Palantir, https\://www\.theguardian.com/us-news/ng-interactive/2025/sep/22/ice-palantir-data

29. INTERVIEW \- Former UK military chief defends Palantir as concerns, https\://aa.com.tr/en/features/interview-former-uk-military-chief-defends-palantir-as-concerns-mount-over-britains-reliance-on-us-tech-/4098487

30. Patient privacy fears as US spy tech firm Palantir wins £330m NHS, https\://www\.theguardian.com/society/2023/nov/21/patient-privacy-fears-us-spy-tech-firm-palantir-wins-nhs-contract

31. Briefing: Concerns Regarding Palantir Technologies and NHS Data, https\://www\.medact.org/2026/resources/briefings/briefing-palantir-fdp/