Google is losing researchers at Google DeepMind over its military AI work. Several staff at the lab have resigned or are in the process of leaving over the company’s contracts with the US and Israeli militaries, the BBC reported on Wednesday, and two of them have spoken on the record despite contracts that restrict them from criticising the company.
Sofia Liguori, a London-based research engineer who spent three years at DeepMind, resigned this week. “Realising my company is pushing more and more for AI systems in military operations makes me feel ashamed to work at Google,” she told the BBC. “I don’t agree with this direction and this has led me to lose faith in my company being true to its stated ethical values.”
Two Researchers Speak On The Record
Liguori joined DeepMind in 2023 and worked on AI systems for routing electricity through power grids more efficiently. She told the BBC she spoke out even though her contract bars public criticism: “In principle I’m not allowed to criticise Google for the duration of my life,” she said, adding that “talking about facts is not criticising.” Inside the company, she said, “people are scared of retaliation and putting their career at risk.” In a separate interview with The Intercept, she described staging a one-person sick-out the day after Google announced its classified Pentagon deal, and co-organising a February demonstration outside the office under banners reading “No AI in the kill chain.”
Andreas Kirsch, an AI researcher who left DeepMind earlier this month, aimed his criticism at oversight rather than the work itself. Google’s governance and controls are “not sufficient or meaningful enough to deploy AI safely and responsibly in military and classified settings,” he told the BBC, “or to develop artificial superintelligence safely in the future.” His sharpest point was about control: “Google seems to have no right to control or veto how the Pentagon uses its AI and also seems to have limited visibility into how it’s being deployed.” Two current DeepMind employees, speaking anonymously, told the BBC that “a lot of people have left the company that feel strongly about this.”
The Classified Deal At The Centre Of The Row
Google is one of several companies that signed contracts with the US military this year to put their AI into classified environments. According to the BBC, Google’s agreement lets its models be used for “any lawful government purpose” and says they “should not be used for domestic mass surveillance or autonomous weapons (including target selection) without appropriate human oversight and control.” OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia and the startup Reflection signed similar contracts, and Defence Secretary Pete Hegseth has said the US will become an “AI-first warfighting force.”
Google defended its process. A spokesperson told the BBC: “We work with public sector organisations worldwide and follow a consistent process for reviewing and entering these contracts, including compliance with our acceptable use policies and AI principles.” The company said it has dedicated channels for employees to raise concerns, from business ethics to misconduct, and “strict workplace policies in place to prevent retaliation of any kind.” The resignations are not DeepMind’s first principled departures this autumn: in September, engineer Robert O’Callahan quit because his team was making AI faster, and DeepMind staff voted to unionise in May.
Why It Matters
The resignations mark a reversal of 2018, when staff pressure pushed Google out of military AI and the dissenters stayed. This time the company is staying in, the contracts are broader — a classified deployment the vendor says it cannot veto or fully observe — and the people raising the strongest objections are the ones walking out. That dynamic matters beyond one lab: every major AI provider has now signed a version of the same classified deal, so the question Kirsch raises, whether a model maker can govern uses it cannot see, applies across the industry at the moment these systems move into war planning. Programmes such as the US Army’s ECHOTRIBBLE agent project show how far autonomous military AI has already moved, and the UN’s first forum on military AI, launched this week, shows regulators are only beginning to build the oversight frameworks that departing researchers say are missing inside the companies.
Background
Google has been here before. In 2018, thousands of employees protested Project Maven, a Pentagon effort to analyse drone footage with AI; some resigned, Google let the contract lapse, and it published its first AI Principles, which listed applications it would not pursue, including weapons. The $1.2 billion Project Nimbus cloud deal with Israel, shared with Amazon, drew fresh staff protests in 2024. In February 2025 Google rewrote its principles, dropping the will-not-pursue list in a post co-written by DeepMind chief Demis Hassabis, while keeping promises of “appropriate human oversight” and alignment with international law and human rights. Jack Shanahan, the retired lieutenant general who led Project Maven, told the BBC the philosophy now seems to be to “go as fast as possible to deploy [AI technology] as widely as possible, potentially at the expense of the rigour and discipline of test and evaluation.” Anthropic’s contrasting approach, setting public damage thresholds in its draft AI safety bill, underlines how differently the leading labs are now answering the same question.
Source: MadRobot, reporting the BBC’s interviews with Sofia Liguori and Andreas Kirsch; BBC News
0 Comments