DoD strikes deals with major tech firms to deploy AI on classified networks

AI classified networks: Pentagon inks $9bn deals with Google, Microsoft, Amazon for secret cloud systems

Sarah Mills | GlobalBeat

The Pentagon signed contracts worth $9 billion with Google, Microsoft and Amazon to deploy artificial intelligence tools on classified networks. Defence officials announced the agreements on Tuesday.

The deals give the three tech giants immediate access to process top-secret data inside tightly secured government cloud environments.

Washington has raced to match China and Russia in military AI applications since generative systems transformed warfare planning in 2023. Previous efforts stalled over security concerns about letting Silicon Valley handle nuclear launch codes, drone feeds and satellite imagery.

Deputy Defence Secretary Kathleen Hicks told reporters the contracts “put cutting-edge algorithms into operators’ hands within months, not years”. She said the AI would scan classified intelligence to spot threats faster than human analysts manage.

Google Cloud CEO Thomas Kurian stood beside Hicks at the Pentagon briefing. He said his firm had built “air-gapped enclaves” that meet the military’s strictest security rules. Microsoft and Amazon issued similar statements confirming their involvement but declined to give contract values.

The arrangement reverses Google’s 2018 pledge to avoid military AI work after employee protests killed its Project Maven contract. Workers had objected to algorithms that helped analyse drone video. The company quietly resumed defence partnerships in 2022 as rivals captured billions in federal spending.

Pentagon officials said the new contracts last five years with options to extend. Each company will run separate AI models so the military can compare results. The systems will digest intercepted communications, radar sweeps and cyberattack patterns to generate threat assessments for commanders.

A senior defence official, speaking on condition of anonymity, said the AI would never control weapons directly. “Human approval remains mandatory for any lethal action,” the official said. Critics remain sceptical.

“This normalises algorithmic warfare,” said Lucy Suchman, an anthropologist who studies military technology at Lancaster University. She warned that even advisory AI systems create pressure to obey machine recommendations during fast-moving crises.

The announcement landed hours before President Donald Trump met defence executives at the White House. Trump posted on Truth Social that America must “dominate the battlefield of tomorrow” through private-sector partnerships. He claimed the deals would “keep our troops safe and our enemies guessing”.

China’s embassy in Washington did not immediately respond to the news. Beijing has invested heavily in military AI, including swarming drones and predictive maintenance for its navy. Russia’s state media dismissed the contracts as “typical Pentagon overspending” but offered no formal comment.

Lockheed Martin and Raytheon shares dipped 2% on fears the tech giants could elbow out traditional defence contractors. Palantir, which already supplies AI to the military, rose 4% as investors bet the Denver firm might win subcontracting work.

The contracts require the companies to store all classified data on government-owned servers inside secure facilities. Each firm must pass quarterly security audits and allow military cyber teams to probe their systems for vulnerabilities. Any data leaks could trigger immediate contract cancellation and criminal charges.

Google’s cloud division has struggled to gain defense market share compared with Microsoft and Amazon, who already hold massive federal cloud contracts. The new AI work gives the Mountain View company a foothold in intelligence operations it has coveted for years. Amazon’s AWS division currently handles the CIA’s commercial cloud, while Microsoft supplies the Pentagon’s everyday email and productivity tools.

Privacy experts questioned whether the systems might ingest Americans’ data swept up during foreign surveillance. The Pentagon official said all AI training would use “lawfully collected information” but declined to elaborate on minimisation rules. Civil liberties groups vowed to press for details through Freedom of Information Act requests.

The Defence Department plans to start testing the AI tools this summer during Pacific military exercises. Commanders will run scenarios involving simulated Chinese missile attacks and North Korean cyber strikes. Officials want algorithms that can stitch together fragmentary radar contacts into coherent threat pictures within seconds.

Each company must hire cleared American citizens to staff the classified cloud centres. Job postings for AI engineers with top-secret clearances have surged 40% since January, according to data from ClearanceJobs.com. Salaries start at $250,000 plus stock bonuses.

Background

The Pentagon began courting Silicon Valley’s AI expertise after AlphaGo defeated the world’s top Go player in 2016. Military leaders realised similar pattern-recognition could revolutionise warfare. Early collaboration attempts stumbled over cultural gaps between hoodie-wearing engineers and uniformed officers accustomed to rigid procurement rules.

Project Maven marked the first major effort to weaponise commercial AI. Google employees revolted when they learned their code might improve drone strike accuracy. Thousands signed protest letters and several engineers resigned. The backlash forced Google to publish AI ethics principles and decline renewal of the Maven contract in 2019. Rivals including Amazon and Microsoft quietly picked up the work, arguing they had a patriotic duty to support national defence.

What’s Next

The Pentagon will select which AI models perform best during the summer Pacific exercises. Officials plan to award follow-on production contracts worth up to $50 billion by December. Each company must prove its algorithms can handle real-time battlefield data without hallucinating fake threats that might trigger accidental wars.

The contracts mark a watershed moment for tech giants’ role in national security. Google, once allergic to military work, now competes aggressively for classified AI contracts. Microsoft and Amazon deepen their lock on federal cloud spending. Whether the systems actually improve battlefield decisions remains unproven. Soldiers will test that question when the first algorithms deploy to conflict zones later this year.