DeepMind workers vote to unionize
- Google DeepMind staff in London asked management on May 5 to recognize two unions after an internal vote to organize over military AI work. - The clearest signal is the margin: 98% of participating workers backed unionization, with the push centered on CWU and Unite. - It matters because DeepMind could become the first unionized frontier AI lab as Google expands classified Pentagon AI work.
Google’s most important AI lab now has an internal labor fight on its hands. Staff at DeepMind in London have voted to unionize and formally asked management to recognize the Communication Workers Union and Unite. The trigger looks pretty clear — Google’s widening willingness to let its AI systems be used in military and surveillance settings, including new classified Pentagon work announced over the past week. ### What actually happened? DeepMind employees in the UK sent a letter to management on Tuesday, May 5, asking Google DeepMind to recognize CWU and Unite as their joint representatives. The organizing vote happened in April, and the workers are framing it as a response to how Google is handling defense contracts and military use of AI. ### Why this team? DeepMind is not some side office. It is central to Google’s frontier-model work — the part of the company building the systems that can later end up inside products, cloud services, and now potentially classified government workflows. If this group organizes successfully, the symbolism is huge because it is the research core, not just a support function, pushing back. ### Why now? The timing lines up with Google’s new Pentagon arrangement. On April 28, reports said Google had reached a deal allowing the Defense Department to use its AI on classified networks. Then on May 1, the Pentagon said a group of major tech companies — including Google, Microsoft that made the internal fear concrete. ### What are workers upset about? Basically, the complaint is not just “we dislike defense work.” It is that employees feel they are losing any real say over where their research ends up. Coverage of the organizing push says workers want Google to stop military applications involving surveillance. ### How strong is the vote? Pretty strong. Multiple reports put the internal vote at 98% in favor among participating workers. That does not mean DeepMind is automatically unionized tomorrow — management still has to recognize the unions or the process has to keep moving through UK labor channels — but it does mean this is not a fringe protest by a handful of employees. ### Why is this different from past Google protests? Google has had employee revolts before — most famously around Project Maven years ago. But this move is more structural. A petition can embarrass management for a week. A recognized union can bargain, file grievances, and keep the issue alive long after the news cycle moves on. That is the real escalation here. ### What’s the bigger stakes? The bigger story is that frontier AI is starting to look less like a pure research culture and more like critical infrastructure with labor politics attached. Once these models move into classified networks, workers are not just arguing about abstract ethics anymore — they are arguing about whether the people building the systems get any voice in how state power uses them. ### Bottom line This is a labor story, but it is really an AI-governance story. DeepMind workers are testing whether employee power can still shape what a leading AI lab will and will not build — just as Google is moving deeper into military work.