Alphabet Inc.’s Google has inked a secretive agreement with the U.S. Department of Defense, granting the Pentagon broad access to its Gemini AI models for classified operations. The deal, confirmed by sources familiar with the matter, permits use across “any lawful government purpose” within secure networks. Details remain shrouded in classification. But one thing stands clear: Google now joins OpenAI and Elon Musk’s xAI in fueling America’s military AI push.
Talks heated up in mid-April. The Information first reported negotiations then, citing two insiders who said the pact would embed Gemini in sensitive environments. By April 28, the ink dried. An anonymous Google source told Engadget the contract bars domestic mass surveillance and autonomous weapons lacking human oversight. Yet it strips Google of any veto power over Pentagon choices.
A Google spokesperson told Reuters: “We believe that providing API access to our commercial models, including on Google infrastructure, with industry-standard practices and terms, represents a responsible approach to supporting national security.” The company echoes safeguards against surveillance and weaponry without controls. Fine words. But lawyers note those clauses lack binding force.
Employee Backlash Echoes 2018 Protests
Nearly 600 workers—many from DeepMind and Cloud divisions—fired off an open letter to CEO Sundar Pichai the day before signing. “Human lives are already being lost and civil liberties put at risk at home and abroad from misuses of the technology we are playing a key role in building,” they wrote, per CBS News. “As people working on AI, we know that these systems can centralize power and that they do make mistakes.”
They demand rejection of classified workloads. Pichai signed anyway. This mirrors 2018’s uproar over Project Maven, Google’s drone-analysis contract that 3,100 employees killed. Google pledged no AI for weapons or surveillance. That vow? Quietly dropped from principles in 2025, as Quartz details. Now, history loops.
Anthropic’s saga sharpens the contrast. The startup inked a $200 million deal but balked at stripping safeguards. Pentagon blacklisted it as a supply chain risk, irking President Trump. OpenAI and xAI complied, securing spots. Google followed suit. Refusal carries costs.
The pact amends an existing contract, Google’s public sector arm confirmed to The Information. It deploys models via API on Google infrastructure. Unclassified Gemini agents already aid over three million DoD users. This escalates to secret realms: mission planning, intelligence sifting, targeting support.
Pentagon eyes more. Officials plan secure sites for AI firms to train on classified data, per MIT Technology Review. Accuracy gains for tasks like threat detection. But risks mount. Models err. Power concentrates.
Strategic Gains Amid Ethical Storms
For Google, revenue beckons. Defense contracts ballooned post-2018; Palantir scooped Maven, ballooning to $340 billion on targeting tech alone. Google Cloud eyes similar windfalls. Pentagon spending hits half a trillion yearly, per The Washington Post. AI slices a growing pie.
Competition rages. China looms. U.S. brass push “AI-first” forces amid Iran tensions. OpenAI’s February deal set the pace. xAI followed. Google’s entry locks in the trio.
Critics warn of slippery slopes. X posts buzz with alarm: “AI Surveillance State,” one calls it. Employees fear opacity—classified use hides harms. “The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads,” the letter insists.
Pentagon gets tools for edge. One Palantir engineer boasted targeting scaled from 2,000 analysts to 20. Gemini could amplify that. But errors in war? Catastrophic.
Google bets on guardrails and oversight. History questions their hold. Pichai’s choice prioritizes national security over internal dissent. The AI-defense fusion deepens. Watch for leaks, lawsuits, more letters. The race accelerates.
WebProNews is an iEntry Publication