AI in Defense and National Security

Learning Objectives

• Identify the leading AI defense companies and their primary capabilities
• Explain how AI is changing intelligence analysis, targeting, and autonomous systems
• Understand the ethical and legal debates around lethal autonomous weapons systems

Defense: AI as a Strategic Priority

Defense and national security have become one of the highest-stakes domains for AI deployment. Major military powers — the United States, China, Russia, and others — are investing heavily in AI for intelligence, surveillance, logistics, and increasingly, autonomous weapons systems.

The US Department of Defense has described AI as foundational to maintaining military superiority. The Pentagon's AI budget has grown substantially each year; multiple programs across every military branch integrate AI for mission-critical functions.

The core AI capabilities in defense mirror those in civilian industry — data analysis, pattern recognition, logistics optimization, computer vision — but applied to contexts where the stakes are measured in lives, strategic advantage, and geopolitical stability.

Palantir: AI for Intelligence and Battlefield Operations

Palantir Technologies (founded 2003 by Peter Thiel and others, originally with CIA funding) is the dominant commercial AI platform for US government intelligence and military applications.

Maven Smart System

Maven Smart System is Palantir's AI-powered battlefield intelligence platform, developed from the original Project Maven (a controversial DoD initiative that Google initially contracted and then withdrew from following employee protests).

Maven Smart System integrates:

• Computer vision for imagery analysis: Processing satellite, drone, and sensor imagery to detect, classify, and track objects of interest — vehicles, equipment, personnel, facilities — at a volume and speed impossible for human analysts
• Multi-source intelligence fusion: Combining signals intelligence (SIGINT), imagery intelligence (IMINT), human intelligence (HUMINT), and open-source intelligence into a unified operating picture
• Natural language querying: Allowing analysts to ask questions of intelligence datasets in plain English rather than requiring mastery of complex query interfaces
• Targeting support: Identifying and tracking potential targets, assessing strike effects, and supporting the overall kill chain

Palantir's contract for Maven Smart System represents a major shift from bespoke government software development toward AI-native commercial platforms. Maven has been designated an official program of record, with its contract ceiling raised to nearly $1.3 billion through 2029. In April 2025, NATO adopted Maven Smart System — the first NATO-wide deployment of the platform. In July 2025, Palantir secured a potential $10 billion Army enterprise agreement, cementing its position as the dominant commercial AI platform for US military operations. Palantir's government revenue hit $570 million in Q4 2025 (66% year-over-year growth).

Palantir AIP for Defense

Beyond battlefield intelligence, Palantir's AIP (Artificial Intelligence Platform) is deployed across defense logistics (maintenance scheduling, supply chain optimization), personnel systems, and administrative functions across the military enterprise.

Anduril: The AI-Native Defense Contractor

Anduril Industries (founded 2017 by Palmer Luckey, creator of Oculus VR) is the most prominent of a new generation of defense technology companies — "defense tech" startups that bring Silicon Valley engineering culture to military hardware and software.

Lattice: The AI Autonomy Platform

Anduril's core technology is Lattice — an AI-powered autonomy platform that integrates data from sensors (radar, cameras, acoustic sensors) to provide autonomous awareness, tracking, and response across large geographic areas.

Lattice is deployed on US military bases for perimeter security — autonomous sensor networks that detect, classify, and track potential threats without requiring a human to monitor every sensor feed. The system presents curated alerts and situational awareness to human operators rather than requiring humans to watch raw sensor feeds.

Autonomous Systems

In March 2026, the Army awarded Anduril a $20 billion enterprise contract for AI counter-drone capabilities, selecting Lattice as the common command-and-control platform. Anduril is also building a $1 billion manufacturing facility in Ohio for autonomous weapons systems production.

Anduril builds purpose-designed autonomous platforms:

Ghost 4 — an autonomous quadcopter drone used for surveillance and, in its armed variants, as a loitering munition (a drone that autonomously navigates to a target area and waits for human command to engage)

Roadrunner — an autonomous air-defense system that can autonomously intercept drones and missiles

Fury — an autonomous combat aircraft designed to fly in coordination with crewed fighter jets as an AI wingman

EagleEye (previewed October 2025) — AI-powered helmet/eyewear for soldiers, with approximately 100 units being delivered to the Army in Q2 2026

Omen (unveiled November 2025) — a hover-to-cruise autonomous air vehicle

All Anduril autonomous systems are designed with human-in-the-loop authorization for lethal engagements — a human must explicitly authorize each strike, even if the system autonomously navigates, identifies, and positions for engagement.

Shield AI: AI Pilots for Autonomous Aircraft

Shield AI has developed an AI pilot called Hivemind that can autonomously fly military aircraft — from small quadcopters to F-16-class jets. In 2023, Shield AI's AI pilot flew a simulated dogfight against a human F-16 pilot and won — demonstrating that AI pilots can execute aerial combat maneuvers faster than human pilots can react.

Shield AI's approach: develop AI that can operate aircraft without GPS, communications, or human direction — enabling autonomous operation in GPS-denied and electronic warfare environments where conventional remotely-piloted vehicles would be unable to operate. The company is now valued at $5.6 billion.

Shield AI has expanded rapidly beyond US Air Force contracts: a $198 million Coast Guard IDIQ for V-BAT maritime ISR (V-BAT 5.3 achieved 100% on all Key Performance Parameters), selection by the Indian Army (~$35 million initial contract, with JSW Defence building a $90 million manufacturing facility in Hyderabad), and international sales to the Netherlands, Romania, Indonesia, and Japan.

The Golden Dome: AI-Enabled Missile Defense

Golden Dome is the Trump administration's initiative for a next-generation US missile defense system — an integrated network of sensors, interceptors, and AI-powered tracking and targeting systems designed to defend the continental United States from ballistic missiles, cruise missiles, and hypersonic weapons.

The technical challenge is extreme: intercepting hypersonic missiles (traveling at 5+ times the speed of sound, on unpredictable flight paths) requires AI-powered tracking and prediction systems, as the decision timelines are too short for purely human decision-making.

The program's cost estimate has risen to $185 billion (including additional space capabilities), with 9 companies now building the command-and-control layer (including Lockheed Martin, Northrop Grumman, and Raytheon). A demonstration under ideal conditions is not expected until the end of 2028 at earliest. Congress is demanding comprehensive spending plans and architecture details.

AI's role in Golden Dome:

• Sensor fusion: Integrating data from space-based sensors, ground radar, and airborne platforms to track threats in real time
• Threat discrimination: Distinguishing real warheads from decoys in a missile salvo — a classic AI classification problem in an adversarial environment
• Fire control: Computing intercept solutions and coordinating interceptor launches on millisecond timelines

Autonomous Weapons: The Ethical Debate

The deployment of AI in military contexts raises profound questions that have no easy answers.

The Legal and Ethical Framework

International humanitarian law (the laws of war) requires that attacks distinguish between combatants and civilians, that attacks be proportionate to military advantage, and that precautions be taken to minimize civilian casualties. These principles have been developed for human decision-makers.

The autonomous weapons debate centers on two questions:

1. Can an AI system reliably distinguish combatants from civilians? In the complex, chaotic environments of actual conflict — with civilians present, with adversaries using civilian infrastructure, with degraded sensor data — AI identification systems make errors. Who is legally and morally responsible for those errors?

2. Should life-or-death decisions be made by machines at all? Even if an AI could theoretically make more accurate target discrimination decisions than a human under stress, many argue that delegating lethal force to a machine crosses a moral line regardless of accuracy.

The US DoD's current policy requires meaningful human control over lethal engagements — a human must authorize each lethal action. In practice, this requirement is subject to interpretation as autonomy increases and decision timelines shorten. NDAA Section 1061 now requires the Pentagon to report waivers of DoD Directive 3000.09 (autonomous weapons safeguards) to congressional defense committees.

The AI Company Ethics Divide

The question of whether AI companies should provide technology for military use has created significant divisions. In July 2025, the Pentagon contracted Anthropic, Google, OpenAI, and xAI ($200 million each) for defense AI applications. Negotiations with Anthropic broke down over "any lawful use" contract language that would have permitted deployment for lethal targeting without the meaningful human authorization Anthropic required. In February 2026, the administration directed agencies to cease use of Anthropic technology and designated the company a supply-chain risk — illustrating the tension between AI safety commitments and national security demands.

The Ukraine Conflict as an AI Testbed

The conflict in Ukraine has become an accelerated real-world test for military AI applications:

• Drone production at industrial scale: Ukraine now produces well over 3 million drones annually, with projections of 7 million in 2026. Drones now cause an estimated 70–80% of battlefield casualties
• AI-boosted accuracy: AI has increased FPV drone strike accuracy from 30–50% to approximately 80% — a transformative improvement in lethality per unit
• AI targeting assistance: Ukraine has used AI-powered tools (including Palantir systems) to process intelligence and support targeting decisions
• Electronic warfare and counter-AI: GPS jamming and electronic warfare designed to defeat drone guidance systems have accelerated development of AI systems resilient to these countermeasures
• "Test in Ukraine" initiative (July 2025): Ukraine invited global manufacturers to test weapons in real combat conditions
• Battlefield AI data sharing (March 2026): Ukraine opened its battlefield AI data to allied nations — described as the "first initiative of its kind in the world" — for training autonomous drone AI models
• Fully unmanned operations: In December 2024, Ukraine conducted its first fully unmanned operation near Lyptsi using UGVs and FPV drones with no infantry

The conflict has demonstrated that autonomous systems change the economics of warfare — cheap, mass-produced autonomous drones can attrit expensive conventional military equipment — with significant implications for how militaries will be designed and resourced.

Key Takeaways

• Palantir's Maven Smart System is the leading AI platform for US military intelligence — with a $1.3 billion contract ceiling, NATO adoption, a $10 billion Army enterprise agreement, and program-of-record designation
• Anduril Industries won a $20 billion Army contract for AI counter-drone capabilities (March 2026) and is building a $1 billion Ohio manufacturing facility — its Lattice platform, autonomous aircraft, EagleEye helmet, and Omen air vehicle represent the commercialization of military AI
• Shield AI's Hivemind AI pilot demonstrated that AI can outperform human pilots in controlled aerial combat simulations — with $5.6 billion valuation and international expansion to India, the Netherlands, and Japan
• All major US defense AI deployments are designed with human-in-the-loop authorization — but the Pentagon-Anthropic dispute over "any lawful use" contract language illustrates the tension between AI safety commitments and military demands
• Ukraine has been an accelerated real-world test of military AI — producing 3 million+ drones annually (projecting 7 million in 2026), with AI boosting FPV accuracy from 30–50% to ~80% and drones causing 70–80% of casualties