Google's classified Pentagon Gemini deal draws internal revolt

Google signed the papers at 4 p.m. on Monday.

The Pentagon can now use Gemini for classified workloads, not just experiments on open data. The deal covers AI running inside air-gapped military networks and states that the tools may be used for "any lawful government purpose". It also requires Google to help adjust AI safety settings and filters when the government asks, while making clear that the company has no right to control or veto lawful government decisions.

The exact signing time matters because this was not a routine renewal. It was a deliberate move into the most sensitive tier of military AI work.

600 signatures, more than 20 VPs

The day before the contract was signed, opposition inside Google erupted.

More than 600 employees signed a protest letter. The signatories included more than 20 vice presidents and directors from DeepMind, a rare public challenge to a CEO-level decision from executives that senior.

The letter put the objection bluntly:

"The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads."

The employees raised four core concerns: AI deployed inside military air-gapped networks cannot be externally supervised; broad language such as "all lawful uses" can override safeguards set by the vendor; staff have no oversight rights in classified missions; and Google had already removed its earlier pledge not to build weapons from its AI principles.

That last point is the hinge of the dispute.

Google has been here before

In 2018, Project Maven triggered one of Silicon Valley's defining labor protests. Thousands of Google workers objected to the company's work on AI for drone target recognition at the Defense Department, and Sundar Pichai eventually let the contract go.

Eight years later, the outcome looks very different. Google revised its AI principles, deleted the weapons clause, won multibillion-dollar Defense Department cloud work, and had already brought Gemini to military users. Extending Gemini into classified workloads simply pushes open the last door.

As one employee organizer put it, "the issue remains unresolved." Time dulled the conflict; it did not settle it.

Anthropic left the room, Google stepped in

The industry context makes the move sharper.

Anthropic had also drawn Pentagon interest, but refused to relax safety settings in the way the government wanted. The result was placement on a military supply-chain risk list, effectively a form of exclusion. CocoLoop covered that earlier dispute.

Once Anthropic stepped back, the space remained open. Microsoft and OpenAI already have classified contracts in operation, and Google has now taken this lane as well. Among the three leading U.S. frontier AI labs, Anthropic is the only one still unwilling to tune safety policy for the military. That is not an accident. It is market sorting: vendors that cooperate win large contracts, while those that refuse are pushed out.

The DeepMind directors who signed the protest appear to understand the symbolism. This is not only a contract. It marks the end of Google's remaining distance from military AI use.

Can the backlash change anything?

Realistically, the 2018 script is unlikely to repeat.

Pichai is no longer facing only a public-relations problem. He is facing geopolitical pressure. The U.S. government wants AI compute and frontier models folded into military capability, and companies that decline risk losing access. Anthropic has already shown what exclusion can look like.

The employee letter spent a day in headlines. Google shares did not suffer, the Defense Department got what it wanted, the CEO secured the contract, and employees returned to work.

What Gemini will do inside the Pentagon's air-gapped classified facilities is a different question. Under the contract language, outsiders have no right to know.

Sources: CocoLoop, Google's updated Pentagon deal uses Gemini for 'any lawful government purpose' with classified data (9to5Google); Google Allows Pentagon to Use Its AI in Classified Military Work (Bloomberg); Hundreds Of Google Employees Revolt Over Pentagon AI Deal (Yahoo News); Google Pentagon AI Deal Expands Into Classified Work (WinBuzzer)