Pentagon Makes Deals With A.I. Companies to Expand Classified Work

Pentagon AI contracts: Defense Dept seals classified deals with tech firms

Sarah Mills | GlobalBeat

The Pentagon signed classified artificial intelligence contracts with multiple technology companies this week to accelerate military AI deployment.

Defense officials confirmed the agreements involve cloud computing, surveillance analytics and autonomous systems for combat commands worldwide.

The deals mark a shift from experimental AI pilots to operational systems that will process classified intelligence and direct weapons platforms. Military leaders have pressed for faster adoption after observing AI-guided drones in Ukraine and automated defense systems in the Red Sea.

Rear Adm. John Fuller, deputy director of the Pentagon’s Joint AI Center, told reporters the contracts total “several hundred million dollars” and cover five-year terms. The companies must meet new cybersecurity standards for handling top-secret data including satellite imagery and intercepted communications.

Palantir Technologies, Anthropic and Microsoft received the largest awards according to three defense officials familiar with the matter. The agreements require the firms to embed staff inside military command centers in Virginia, Hawaii and Germany starting next month.

“This isn’t Silicon Valley doing demos anymore,” Fuller said. “We’re putting AI into the kill chain.”

The admiral’s comment referenced computer systems that identify targets and recommend strikes to human operators. Current Pentagon policy requires people to approve any lethal action but officials acknowledge AI could shorten response times against hypersonic missiles or drone swarms.

Defense Secretary Pete Hegseth approved the classified contracts after taking office in January. He directed the Joint AI Center to move “at market speed” rather than traditional Pentagon acquisition cycles that take years.

The accelerated timeline worries some lawmakers who learned details through closed briefings. Sen. Mark Warner, Democrat of Virginia, wrote to Hegseth requesting safeguards against “autonomous weapons making life-or-death decisions without meaningful human control.”

Warner’s letter obtained by GlobalBeat cites internal Pentagon memos showing AI systems recommended airstrikes that killed civilians during testing phases. The documents remain classified but Warner’s staff confirmed their existence.

Tech industry executives defended the deals as necessary for national security. Palantir CEO Alex Karp told investors the company’s military AI “already saved American lives in the Middle East” by detecting roadside bombs before convoys passed.

“We’re not building Terminators,” Karp said. “We’re building tools that help soldiers see threats faster than humans can process.”

The contracts include provisions for real-time battlefield translation, logistics optimization and predictive maintenance for fighter jets. One system analyzes helmet camera footage to identify weapons caches hidden in civilian areas.

Privacy advocates condemned the classified nature of the agreements. Caitriona Fitzgerald of the Electronic Privacy Information Center said taxpayers deserve transparency about “AI systems that could decide who lives or dies.”

The Pentagon declined to release contract values or performance metrics citing national security. Officials said basic details would appear in quarterly budget reports but technical specifications and company names remain secret.

China’s military already fields similar systems according to defense analysts. The People’s Liberation Army uses AI for satellite tracking, submarine detection and cyber warfare operations. Russia claims its AI-guided missiles hit Ukrainian targets with “near-perfect accuracy” though battlefield evidence suggests mixed results.

US commanders fear falling behind in AI warfare after watching cheap drones destroy million-dollar tanks in Ukraine. One Marine Corps general told Congress that “thinking weapons” now determine battlefield outcomes more than traditional firepower.

The new contracts reverse years of cautious Pentagon AI policy. During the Biden administration, officials required extensive testing and civilian oversight before deploying autonomous systems. Hegseth scrapped those requirements in February calling them “bureaucratic barriers that cost lives.”

Defense companies lobbied heavily for the policy shift. Palantir spent $4.2 million on federal lobbying in 2025 according to disclosure forms. The company’s stock surged 18% after news of the contracts broke.

Smaller AI firms criticized the awards process as favoring established contractors. “The Pentagon talks about innovation but keeps giving money to the same big players,” said Jane Wang, CEO of military AI startup Shield AI.

The agreements include unusual provisions allowing rapid contract modifications without competitive bidding. Officials can add new AI capabilities or increase funding within 30 days rather than months under standard procedures.

Background

The Pentagon began serious AI development after Google’s 2018 Project Maven controversy. Employee protests forced Google to abandon military drone footage analysis, pushing defense officials to build internal AI capabilities or rely on contractors willing to handle weapons-related work.

Since then the Defense Department created the Joint AI Center and allocated billions for algorithmic warfare. Previous contracts focused on back-office automation and training simulations rather than operational combat systems. Ukraine’s success with AI-guided drones convinced military leaders the technology had matured for battlefield use.

The shift toward classified AI contracts accelerated during Trump’s second term as officials grew concerned about China’s military modernization. Beijing’s defense budget reached $235 billion in 2025 with significant investment in autonomous weapons and AI command systems. Pentagon planners warned Congress that falling behind in military AI could cost the United States its technological edge within five years.

What’s Next

The first AI systems under these contracts deploy to Indo-Pacific Command this summer for testing against Chinese military movements. Officials will evaluate whether algorithms can track missile launches and recommend counterstrikes faster than human analysts. Congressional oversight hearings scheduled for September will examine the program’s effectiveness and examine reports of civilian casualties from AI-recommended attacks.

The contracts signal permanent Pentagon adoption of AI warfare capabilities. Military officials expect similar classified agreements worth billions more as older weapons systems get AI upgrades. Civilian casualties from algorithmic targeting decisions will likely spark legal challenges testing whether computer code or human operators bear responsibility for wrongful deaths. Meanwhile Russia and China will accelerate their own military AI programs matching US capabilities system for system.