Google Joins Pentagon's Classified AI Ecosystem, Reversing 2018 Ethics Stance

Google has quietly signed a deal to give the US military access to its artificial intelligence models for classified work. The agreement, reported by The Information on April 28, 2026, allows the Pentagon to use Google's Gemini AI for "any lawful government purpose" on its classified networks. That phrase matters: classified networks handle mission planning and weapons targeting.

Google is not alone. The company joins OpenAI and Elon Musk's xAI, which have already signed similar agreements with the US Department of Defense (now officially renamed the Department of War by President Donald Trump). Every major commercial AI lab in America now sits inside the Pentagon's classified network.

The Deal Google Had to Agree To

This isn't just Google providing a cloud service. The agreement explicitly requires Google to help adjust its AI safety settings and filters at the US government's request, meaning the guardrails Google built into its AI systems can be loosened or changed, on demand, by the Pentagon.

The contract does include some stated limits. The agreement says the AI "should not be used for domestic mass surveillance or autonomous weapons (including target selection) without appropriate human oversight." That sounds protective. But the same contract immediately undercuts it, adding that the agreement "does not confer any right to control or veto lawful Government operational decision-making."

In other words, Google agreed to limits it cannot actually enforce.

A Reversal Eight Years in the Making

In 2018, Google's employees successfully pressured the company to exit the Pentagon's Project Maven initiative, a program using AI to analyze drone footage. Google pulled out entirely, positioning itself as a tech company with ethical lines it would not cross.

That position lasted eight years.

Now 580+ Google employees, including more than 20 directors and VPs, have signed an open letter to CEO Sundar Pichai asking him to refuse this new classified military deal. Unlike in 2018, the opposition has not worked.

What changed? The Pentagon made compliance the price of admission. Secretary of Defense Pete Hegseth issued a memo in January 2026 requiring all Defense Department AI contracts to include "any lawful use" language. The administration then designated Anthropic, the one major AI lab that refused to strip its safety guardrails on autonomous weapons, a "supply-chain risk" and directed federal agencies away from its products. The message to remaining labs: comply, or lose government business worth hundreds of millions of dollars.

The Contract Numbers

The Pentagon signed contracts worth up to US$200 million each with Google, OpenAI, Anthropic, and xAI in 2025. OpenAI's classified deal alone is estimated to be worth between US$500 million and US$2 billion over five years. For AI companies burning cash at scale, defense contracts represent a category of revenue that is hard to walk away from.

What This Means for Asian Enterprises

For executives in Asia managing enterprise AI decisions, this story raises questions that go beyond US domestic politics.

Indonesia, Malaysia Block AI Chatbot Over Deepfake Abuse

Indonesia and Malaysia blocked Grok after it was repeatedly used to generate nonconsensual sexual deepfakes. A regulatory turning point for APAC marketing leaders navigating AI content risks.

Mission MediaMission Media

The AI tools powering your marketing automation, content workflows, and customer analytics are products of the same companies now operating under classified US military agreements that allow their standard safety policies to be adjusted at government request.

This does not mean the tools you use today are suddenly dangerous. But it raises a legitimate question about AI vendor reliability. When an AI company's published safety commitments can be renegotiated with sufficiently powerful clients, how stable are those commitments as a supplier guarantee?

The gap between what an AI vendor publicly promises and what it contractually delivers to its largest clients is now, for the first time, visible and documented. Google has not publicly commented on the deal's terms, and neither Alphabet nor the Department of War responded to requests for comment.

What is confirmed: the deal exists, Google signed it, and the classified AI ecosystem now includes every major US lab.