May 5, 2026
Eight years ago, a protest inside Google’s walls changed the course of the company’s relationship with the U.S. military. Nearly 4,000 employees signed a petition opposing “Project Maven,” a Pentagon contract to use AI for analyzing drone footage. Twelve engineers resigned. The backlash was so fierce that Google let the contract expire and publicly committed never to build AI for weapons or surveillance. It felt, at the time, like a landmark moment — proof that the conscience of Big Tech could act as a check on the military-industrial complex.
That moment now looks very far away.
Last week, Fortune and The Information confirmed that Google has signed a new classified contract with the U.S. Department of Defense, one that allows the Pentagon to deploy its Gemini AI models for “any lawful governmental purpose” — including, reportedly, inside classified networks used for mission planning and weapons targeting. The deal is an amendment to an existing $200 million contract with the Pentagon, but the new layer goes far deeper: into the kind of secure, compartmentalized systems that handle some of the military’s most sensitive operations.
More than 600 Google and Google DeepMind employees have signed an open letter to CEO Sundar Pichai urging the company to reverse course. One DeepMind research scientist, Andreas Kirsch, said publicly that he was “incredibly ashamed” and “speechless.” The employees wrote that they believe “Google should not be in the business of war” and that they want AI to “benefit humanity, not be used in inhumane or extremely harmful ways.”
It is the echo of Project Maven. But the outcome may be very different this time.
A Changed Landscape
The AI industry has undergone a seismic shift since 2018. What was once a field driven primarily by academic researchers and idealistic engineers is now one of the most strategically contested sectors in the global economy. The U.S. government has made AI a national security priority, and the pressure on tech companies to participate — or be left out — has intensified enormously.
The stakes became clearer just weeks before the Google news broke. The Pentagon publicly designated Anthropic a “supply chain risk” and effectively barred it from classified government work. The reason? CEO Dario Amodei had refused to agree to terms allowing unrestricted use of its Claude models for military purposes. The message was blunt: companies that won’t cooperate with the government on AI may find themselves excluded from some of the largest and most strategically important contracts in the country.
Against that backdrop, Google’s decision looks less like an about-face and more like a calculated survival move. OpenAI has agreed to similar classified-use terms. So has xAI, Elon Musk’s AI company. Nvidia, Microsoft, and Amazon have all signed comparable agreements. The industry consensus, it appears, has shifted.
Why Employee Leverage Has Weakened
In 2018, Google’s workforce was more cohesive, more activist, and more capable of exerting meaningful internal pressure. The company was earlier in its AI journey, less dependent on government contracts, and more sensitive to reputational risk in a talent market where the best researchers had many other options.
Today, the landscape for employee activism inside Big Tech looks quite different. A wave of tech layoffs between 2022 and 2024 reshaped the power dynamic between workers and management. Signing a protest letter carries risk in a way it didn’t before. And for a company like Google, which is locked in existential competition with OpenAI, Microsoft, and a growing array of well-funded AI startups, appearing to hamper its own growth by refusing government contracts is a harder position to defend in boardrooms and investor calls.
“The leverage has just changed,” one industry analyst told Fortune. “In 2018, AI talent was scarce and Google could afford to listen to its employees. Now, the company is fighting for its AI future on multiple fronts simultaneously.”
That doesn’t mean the employee concerns are less legitimate. If anything, the moral stakes are higher. AI models in 2026 are vastly more capable than what existed during Project Maven. An AI system that can plan, reason, and adapt — deployed in classified military contexts with limited public oversight — raises questions that a simple drone-footage classifier did not.
What “Any Lawful Purpose” Actually Means
The phrase at the heart of the controversy — “any lawful governmental purpose” — is doing a lot of work. Government lawyers have considerable latitude in defining what is “lawful” in classified contexts, and the public has little insight into what specific applications the Pentagon is actually building.
Critics argue that the clause is essentially a blank check. The AI systems being deployed are not narrow tools with defined functions; they are general-purpose models capable of being adapted to a vast range of tasks. Unlike a weapons system that requires years of development and testing, an AI model can be repurposed and redeployed rapidly, with minimal external oversight.
Google has said the partnership includes “strict data protection, governance and human oversight” safeguards, but the classified nature of the contract makes independent verification impossible.
A New Normal for AI and Defense
The Google Pentagon deal is not an anomaly — it is increasingly the norm. The U.S. government is aggressively building out AI capabilities across defense, intelligence, and national security, and it is doing so by pulling the biggest AI companies into classified work on terms that prioritize operational flexibility over public transparency.
What is striking is how quickly the field has moved. Not long ago, the idea of frontier AI models — the most powerful and capable systems ever built — being deployed in classified military networks for unspecified purposes would have seemed alarming enough to spark widespread public debate. Today, it barely registers beyond the tech press and a few hundred outraged engineers.
The Project Maven revolt forced a genuine reckoning inside Google. Whether the 2026 version of that revolt will have any similar effect remains deeply uncertain. What is clear is that the industry has crossed a threshold, and the ethical frameworks that once seemed adequate to govern it are struggling to keep pace.
For the employees who signed that letter to Sundar Pichai, the question isn’t just whether Google will listen. It’s whether, in the current environment, anyone in a position of power is listening at all.
Sources: Fortune, The Information, Euronews, Washington Post, eWeek, Inc., IBTimes UK
