Google’s Pentagon AI Deal Was Signed While Its Own Employees Were Signing a Petition Against It

Google’s recent $200 million AI contract with the Pentagon provided one of the clearer examples of this type of conflict that only occurs inside very large organizations. Nearly at the same time that over 600 workers were discreetly circulating an open letter calling for the agreement to be terminated, the company’s leadership was signing the final contract that the workers were attempting to thwart.

The two occurrences took place concurrently, in separate buildings, sometimes on different levels of the same buildings, and most likely with at least some individuals in the middle who were aware of both. The outcome was the kind of introspective moment that causes awkward morning standups for weeks.

The contract is important in and of itself. The Pentagon will now be able to utilize Google’s Gemini family of AI models on classified networks for “any lawful government purpose,” a phrase that has concerned both the personnel base and legal analysts. When interpreted carefully, it phrase is harmless. When interpreted broadly,

it might include everything from threat analysis to logistical planning to capabilities that the general public would find more difficult to comprehend. Speaking with defense analysts who have examined comparable contracts, it seems that the language’s breadth was deliberate. Procurement officers at the Pentagon typically advocate for maximal flexibility. Tech companies typically resist until the sum of money becomes unavoidable.

The employee’s letter to CEO Sundar Pichai was direct. Google’s reputation would suffer “irreparable damage” if the transaction went through, the signatories said. They cited the company’s declared AI values. They made reference to particular technologies that engineers had developed with civilian use cases in mind; these technologies could theoretically be used in defense situations, which the engineers themselves found unsettling.

Directors, vice presidents, and a sizable delegation from Google DeepMind, the company’s AI research division, which has historically produced some of the most vocal researchers on matters of military applicability, were among the signatories.

The timing was what gave the occasion an almost unreal quality. While drafting their objections, a number of employees who signed the petition told reporters they were unaware that the sale was being finalized. According to reports, internal communications regarding the contract had been restricted to those who needed to know and did not reach the engineering levels.

The contract had progressed past the stage where employee pressure could actually reverse it by the time the open letter became widespread enough for leadership to notice it. It’s important to consider if that sequencing was intentional or just a result of how big businesses operate.

It is impossible to ignore the historical resonance. After hundreds of employees signed a similar petition and a few left in protest, Google withdrew from the Pentagon’s previous AI imagery program, Project Maven, in 2018.

At the time, many saw that incident as a turning point in which tech workers demonstrated their ability to put pressure on leadership about military involvement. The same trend is still occurring eight years later, but the results appear to be different. The leadership did not concede. The agreement was signed. Although there is a petition, it seems that employees’ cultural influence has significantly diminished.

It’s difficult to ignore the larger industrial change that underlies this. Just a year ago, OpenAI claimed to be wary of defense alliances, but it has now developed stronger links with the Pentagon. With its Azure government cloud, Microsoft has been there for many years. Even smaller AI companies are now vying for Pentagon contracts that would have been deemed politically toxic in 2020.

The industry’s framing has evolved. Nowadays, a lot of executives contend that frontier AI capabilities—regardless of who supplies them—will be utilized for defense applications, and that American companies remaining out just gives the advantage to others, particularly foreign rivals. It is debatable if that argument can withstand more ethical scrutiny, but it has undoubtedly taken center stage in tech leadership circles.

Compared to the contract value, the longer-term cost is more difficult for Google to quantify. Internal trust takes weeks to lose and years to create. Many developers considered staying at Google for the remainder of their careers after the 2018 Project Maven moment. A similar generational shift in attitude could result from the 2026 incident, in which the leadership signed while workers were publicly protesting.

Speaking with current Google employees in the days following the announcement has given me the impression that a basic aspect of the staff-executive leadership relationship has been damaged. Perhaps not damaged. However, the bruises don’t heal on their own.

Whether Google can handle the dissatisfaction without losing too many of its important AI researchers—who have lots of options at rival companies and well-funded startups—will determine what happens next. Because the work is fascinating and the pay is unparalleled, some will stick around. Others will move on in silence. For upcoming historians of the AI era, the petition itself is now a part of the public record. And despite the opinions of 600 of Google’s most seasoned engineers, the Pentagon, with a recently inked $200 million deal, has received exactly what it wanted.