DeepMind workers unionize to block military AI

You might think frontier AI researchers only care about scaling compute, token throughput, or benchmark scores. At Google DeepMind's London headquarters, researchers are drawing a completely different line in the sand.

On Tuesday, staff at the lab demanded that Google management formally recognize the Communication Workers Union (CWU) and Unite the Union as their collective bargaining representatives. Around 98 percent of CWU members at DeepMind supported moving forward. If successful, the move will cover at least 1,000 employees based in London.

Google management now has 10 working days to voluntarily grant union recognition before staff trigger formal legal proceedings to force it.

Pushing back on the military pipeline

The driving force behind the union drive isn't free snacks or remote work policies. It is Google's growing involvement in defense and military AI contracts.

Staff want legally binding guardrails. Their list of demands includes a clear commitment from Google not to develop weapons, surveillance systems, or harmful technology; mandatory bargaining over how AI deployments affect employee workloads and job security; and the contractual right to abstain from projects that violate their personal moral or ethical standards.

DeepMind workers globally are also reportedly considering in-person protests and "research strikes," where they would refuse to build improvements for flagship products like Gemini.

Google is already trying to cool the room. Company spokesperson Kristen Morea said management received the letter, emphasizing that at this stage in the process, no formal vote to unionize has taken place, and noted that Google remains focused on constructive dialogue.

Old tensions, new leverage

This confrontation didn't come out of nowhere.

Just days earlier, hundreds of Google workers signed an open letter urging CEO Sundar Pichai to reject classified Pentagon AI contracts. Google, alongside OpenAI and Nvidia, reportedly signed agreements allowing the US Department of Defense to deploy their models for any lawful government purpose. Back in 2024, Google fired more than 50 employees following protests over its contracts with the Israeli government.

CWU national officer John Chadfield noted that frontier AI researchers are now connecting with traditional labor organizing to push back against defense-contract drift.

Why it matters

We think this union drive is a clever and consequential escalation. Firing dozens of individual protesters might work once, but negotiating with an officially organized union of 1,000 elite AI researchers is a completely different operational headache for Google.

For tech workers and researchers, this sets a massive precedent. It moves ethical dissent out of Slack channels and HR complaint boxes and turns it into a protected labor negotiation. If successful, having the formal right to refuse working on military deployments could become a baseline expectation for top-tier talent across Silicon Valley and Europe.

For the AI industry, it exposes a major bottleneck in the race for government defense contracts. Big tech executives are rushing to sign massive military deals, but the underlying models require constant maintenance and algorithmic breakthroughs from a relatively small pool of specialized minds. If DeepMind staff actually follow through with research strikes on Gemini, Google's enterprise and defense roadmaps could take a direct hit.