Union negotiations between Google DeepMind and its London-based staff hit an early wall this week, after an opening arbitrated session on Wednesday was attended only by HR representatives rather than senior DeepMind leadership. The Communication Workers Union, one of two unions seeking recognition alongside Unite the Union, said the absence of executives signalled bad faith and that talks had stalled at the first meeting. DeepMind disputes that characterisation.
DeepMind employees asked Google to recognise the CWU and Unite as joint representatives in May 2026. Google denied that request but agreed to participate in a third-party arbitrated process. Wednesday's session brought together union officers, DeepMind staff pushing for recognition, the arbitrator, and DeepMind HR — but no senior management from the lab itself.
John Chadfield, the CWU officer who attended, said the composition of the meeting told union representatives everything they needed to know about Google's posture. Al Verney, a Google DeepMind spokesperson, countered that the opening session is a scoping exercise and that the right people were in the room for that stage.
Key facts
- 01DeepMind's London staff asked Google in May 2026 to recognize the CWU and Unite the Union as joint representatives; Google declined but agreed to third-party arbitrated talks.
- 02The first negotiation session on July 1, 2026 was attended by HR representatives, not senior DeepMind leadership, prompting CWU to say talks have stalled.
- 03The unionization push began in February 2025 after Alphabet dropped its pledge against using AI for weapons and surveillance.
- 04About 600 US-based Google employees signed a letter protesting an April 2026 Pentagon deal allowing the DoD to use Google AI for 'any lawful government purpose.'
- 05The DoD has confirmed AI deals with seven leading labs, including Google, SpaceX, OpenAI, and Microsoft, for use on classified networks.
During the meeting, a DeepMind employee read a prepared statement on behalf of colleagues supporting unionisation. According to multiple people with knowledge of the meeting, HR representatives interrupted the reading twice. The letter accused Google of routing employee concerns to HR rather than engaging directly, of reconfiguring internal chat channels used to organise, and of blocking staff from replying to company-wide communications about the union bid.
The letter alleged that employees who tried to work around those restrictions were reprimanded. A DeepMind employee involved in drafting it described the pattern as a deliberate intimidation strategy, using the phrase 'union-busting techniques' and asking to remain anonymous. Verney said Google will continue to engage constructively in the arbitrated process and offers employees other channels to raise concerns.
The organising drive traces back to February 2025, when Alphabet removed language from its AI ethics guidelines pledging not to develop AI for weapons or surveillance. That policy shift is what pulled a critical mass of DeepMind researchers into the union conversation in the first place — the ethics commitments had been part of the pitch that recruited many of them.
The stakes have widened since. In late February 2026, staff at DeepMind and OpenAI signed an open letter backing Anthropic after the US Department of Defense sought to designate the lab a supply-chain risk over its refusal to allow its models to be used in autonomous weapons or mass surveillance. In April 2026, a deal surfaced allowing the Pentagon to use Google AI for 'any lawful government purpose.' Roughly 600 US-based Google employees signed a letter protesting the terms.
The Department of Defense has since confirmed deals with seven leading AI companies — Google, SpaceX, OpenAI, and Microsoft among them — to run their models on classified networks. Jenn Crider, a Google spokeswoman, has defended the arrangement, saying the company remains committed to a private- and public-sector consensus that AI should not be used for domestic mass surveillance or autonomous weaponry without appropriate human oversight.
The London push is not the first attempt by Google workers to organise. In 2021, US-based Google staff formed the Alphabet Workers Union, which Alphabet does not recognise for collective bargaining but which has secured agreements on behalf of Google contractors. A recognised bargaining unit inside DeepMind's London office would go further, giving frontier-lab researchers a formal seat at the table on questions like military contracts and model deployment policy.
If negotiations do not progress, Chadfield said, the unions will ask an arbitration committee to compel Google to recognise them. He framed the current impasse as one where his side wants an amicable settlement but is not seeing reciprocal movement from Google.
For AI companies, the DeepMind negotiation is the first serious test of whether frontier-lab researchers — the highest-leverage employees at the highest-leverage companies in the sector — can convert their moral objections to defence work into contractual leverage. If the London arbitration ends in recognition, expect similar drives to surface quickly at OpenAI, Anthropic, and Microsoft's AI units, all of which now sit on the same DoD roster. If it ends with Google outlasting the process, the message to every other lab will be that the classified-networks business is worth absorbing the internal cost.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
