More than 600 Google employees have broken ranks with their own leadership, demanding the company abandon a newly struck deal that allows the US Department of Defense to deploy the firm’s most powerful artificial intelligence models in classified military operations. The open letter, addressed to chief executive Sundar Pichai and signed by staff across Google DeepMind, Google Cloud and other units – including over 20 directors, senior directors and vice presidents – rips through the carefully managed language of corporate responsibility. It warns that classified work ‘lacks transparency by nature’ and offers no safeguard against the use of their creations for ‘sorting individuals or targeting innocent civilians’. The revolt, organised in apparent haste after the deal’s existence leaked, marks one of the most significant internal challenges to a technology giant’s embrace of the national security state since the 2018 employee uprising that forced Google to abandon Project Maven, an earlier Pentagon drone-imagery contract.

Viewed from Washington, the agreement is a straightforward procurement necessity. The Department of Defense has raced to harden its classified networks with generative AI, signing parallel accords with OpenAI and Elon Musk’s xAI, each worth up to $200 million. These tools are intended for mission planning and intelligence analysis, not autonomous targeting, the Pentagon insists; the contract text stipulates that any AI system ‘is not designed or intended to be used for mass domestic surveillance or for autonomous weapons without appropriate human oversight and control’. Under the terms, Google retains no veto power over lawful government operational decisions. Company spokespeople, speaking from Mountain View, call the agreement an amendment to an existing support contract and stress that Google assists US authorities in both classified and unclassified capacities, framing the work as a patriotic duty rather than a moral rupture.

Yet from London to Milan, that framing is meeting fierce resistance where it lands closest to home: inside the company’s own research cadres. Andreas Kirsch, a research scientist at Google DeepMind, posted on social media that he was ‘incredibly ashamed’ and ‘speechless’ at the deal. The internal letter, whose signatories now number upwards of 560 in some counts, argues that AI should ‘benefit humanity’ rather than be employed in ‘inhumane or deeply harmful ways’, explicitly citing lethal autonomous weapons and mass surveillance. Arab-language outlets framing the story for readers in the Gulf and Levant have seized on this ethical schism, foregrounding the workers’ warning that their technology risks ‘undermining civil liberties’ – a phrase that resonates powerfully in regions where Western military technologies have long cast a shadow over civilian populations. Russian business dailies, for their part, have reported the deal with a dry, almost technical curiosity, noting the restriction against autonomous targeting while underscoring that Google cannot control state decisions – a detail that, in Moscow’s reading, seems to underline Washington’s determination to harness Silicon Valley without yielding sovereignty.

What makes this confrontation more consequential than a corporate squabble is the quiet collapse of the boundary Google once loudly proclaimed. The company’s 2018 AI principles promised it would not design weapons or pursue technologies whose principal purpose is to cause harm. The new classified contract may not officially breach that pledge – human oversight is nominally preserved, and the work is compartmentalised away from consumer products – but it erodes the spirit of a firewall that many researchers believed was permanent. Analysts in European capitals note that the episode exposes a structural tension: Western democracies want to outpace authoritarian rivals in military AI while retaining the moral high ground, yet the engineers most capable of delivering that edge are increasingly unwilling to be drafted into opaque state projects. As the NATO alliance deepens its own reliance on algorithmic warfare, Google’s internal rebellion suggests the real front line in the AI arms race may not be in a distant data centre, but in the conscience of the people building it.