Google signs classified AI deal with Pentagon despite pushback

Silicon Valley's military AI divide is turning into a full-scale corporate sprint.

On Friday, the U.S. Department of Defense announced that it has signed deals with Nvidia, Microsoft, Amazon Web Services, and Reflection AI to deploy their AI hardware and models on classified military networks for "lawful operational use." The announcement follows previous agreements already secured with Google, SpaceX, and OpenAI.

Meanwhile, the pressure inside Google just hit a boiling point. More than 600 Google employees signed an internal letter to CEO Sundar Pichai demanding the company block the Pentagon from deploying its AI in classified settings, according to The Washington Post. The signatories reportedly include dozens from Google DeepMind, alongside more than 20 principals, directors, and vice presidents.

Diversifying around Anthropic

The Pentagon's aggressive expansion across multiple tech giants isn't random. It follows an escalating legal fight with Anthropic over usage restrictions.

The Defense Department wanted unrestricted access to Anthropic's models, while the AI startup insisted on strict guardrails banning its models from being used for domestic mass surveillance and fully autonomous weapons. When talks collapsed, the Pentagon attempted to designate Anthropic a "supply-chain risk," though Anthropic won an injunction against the label in March.

Now, the Pentagon is making sure it never depends on a single vendor again. DOD officials explicitly stated that the goal is to build an architecture that "prevents AI vendor lock-in" and guarantees long-term operational flexibility. The new deployments are heading to Impact Level 6 (IL6) and Impact Level 7 (IL7) networks, the military's highest security tiers requiring physical isolation and strict audit controls for critical national security workloads.

Outside classified networks, adoption is already broad. The Pentagon noted that over 1.3 million DOD personnel currently use GenAI.mil, an enterprise portal built for non-classified tasks like document drafting, research, and basic data analysis.

Pushback meets defense budgets

The Google letter warns that classified contracts create a black box where models can be weaponized out of sight. "The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads," the letter reads. "Otherwise, such uses may occur without our knowledge or the power to stop them."

Yet with OpenAI renegotiating terms back in February and Microsoft already entrenched in defense networks, Google chose not to sit on the sidelines, despite open revolt from some of its most senior AI researchers.

Why it matters

For enterprise users and developers, this deal draws a stark dividing line across the AI industry. Anthropic chose a high-stakes courtroom showdown over softening its guardrails, while Google, OpenAI, and Microsoft decided the business of national defense is too big to abandon. Which ecosystem developers back now carries clear philosophical baggage.

For defense tech, the Pentagon's playbook is ruthless and effective. By locking in Nvidia's silicon, the dominant cloud giants, and multiple frontier model builders simultaneously, the military avoids being held hostage by any single lab's safety charter. In doing so, it creates intense commercial pressure on labs to dial back restrictive policies if they want federal dollars.

We think Google leadership may have secured a seat at the Pentagon table, but the internal fallout could be steep. DeepMind's brand was built on high-minded safety and research purity. Forcing classified defense deployments over the objections of hundreds of core scientists risks alienating the very talent Google needs to win the frontier AI race in the first place.