A growing wave of internal dissent is roiling Google DeepMind, Google’s leading artificial intelligence subsidiary, as multiple senior AI researchers and engineers resign in protest over the company’s expanding partnerships with military entities around the globe. The BBC has confirmed that several current and former employees have stepped away from their roles, with multiple insiders speaking out on condition of anonymity – and one whistleblower even breaking contractual non-disclosure rules to share her concerns publicly.
Sofia Liguori, a London-based research engineer who spent three years at Google DeepMind, submitted her resignation earlier this week, becoming one of the most high-profile critics to exit the firm over the military contracts. Speaking exclusively to the BBC despite contractual restrictions barring her from public comment, Liguori said the company’s growing push to integrate AI into active military operations left her deeply ashamed to remain on staff.
Her departure was driven not just by US military contracts, but also by Google’s expanding work with the Israeli military, she explained. “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,” Liguori said. Her criticism echoes a core tension at the heart of Google’s public commitments: the company’s official AI Principles explicitly require human oversight of all AI systems and mandate alignment with international law and human rights standards. But former and current employees, as well as independent experts, say those commitments are not being honored in classified military work.
Andreas Kirsch, another prominent AI researcher who resigned from Google DeepMind earlier this month, echoed Liguori’s concerns. Kirsch argued that Google’s internal governance and oversight frameworks are far too weak to ensure AI is deployed safely and responsibly in classified military settings – and that the gap also puts long-term AI safety work at risk. Critically, Kirsch noted that Google has no ability to review or veto how the Pentagon uses its AI tools, and only limited visibility into real-world deployment of the technology on the battlefield. Current AI models, he added, are simply not robust enough or adequately trained for the high-stakes context of warfare.
Google’s push into military AI comes as the US Department of Defense has made rapid integration of AI into all branches of the armed forces a top policy priority under the Trump administration. US Defense Secretary Pete Hegseth has publicly stated the administration is committed to transforming the US military into an “AI-first warfighting force”, and the Pentagon has already awarded a wave of contracts to leading tech firms to expand AI use across classified operations. Google was one of multiple major technology companies – alongside OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia, and startup Reflection – to sign military AI contracts earlier this year. These agreements expand AI use across a wide range of military functions, from logistics and transport to operational planning and active battlefield operations.
Google defended its practices in an official statement to the BBC, saying the company follows a consistent, rigorous review process for all public sector contracts, and that all agreements comply with its internal acceptable use policies and published AI principles. The company added that it maintains dedicated internal channels for employees to raise ethical concerns, and enforces strict policies prohibiting retaliation against staff who speak up. The specific 2025 contract Google signed with the Pentagon prohibits use of the company’s AI for domestic mass surveillance and autonomous target selection without formal human oversight, restricting use to “any lawful government purpose”.
Critics, however, warn that these safeguards are insufficient, even on paper. Rapid advances in AI capabilities have outpaced the development of regulatory guardrails and rigorous testing protocols, raising urgent questions about whether the technology is reliable enough for combat use. Retired Lieutenant General Jack Shanahan, the former director of the Pentagon’s landmark Project Maven – the Defense Department’s flagship AI-enabled intelligence and targeting platform – told the BBC that the current Pentagon philosophy prioritizes speed of deployment over rigorous testing and evaluation. Project Maven, developed by defense contractor Palantir with integrated technology from multiple leading AI firms, uses AI to speed up target identification and selection for strikes, cutting down processing time dramatically for intelligence teams.
Pentagon officials defend the rapid push to adopt AI, arguing that speed is the defining factor for victory in the new AI era of warfare. But independent AI ethicists and policy experts warn of catastrophic risks. Sarah Shoker, former head of OpenAI’s Geopolitics team, explained that AI drastically compresses the time human operators have to make life-or-death targeting decisions. When an AI generates hundreds of potential targets in a compressed timeline, it is functionally impossible for a human operator to meaningfully review every output for errors, she said – a gap that drastically increases the risk of civilian casualties.
Multiple Google employees told the BBC that internal culture discourages any public criticism of the company’s military AI work, with many staff fearing career retaliation for speaking out. Liguori noted that her employment contract includes a lifetime ban on criticizing Google, but she chose to speak publicly anyway, arguing that “talking about facts is not criticising” and that the public has a right to know about the company’s activities. One anonymous current Google DeepMind employee said many staff who joined the company to work on ethical AI for public good are now considering leaving, though the company’s strong brand means it will likely have no trouble filling open roles with early-career workers. “It could be a changing of the guard,” the employee said.
Controversy has also flared over Google’s 2024 Project Nimbus, a $1.2 billion joint cloud and AI contract with Amazon to serve the Israeli government and military. Google has previously stated the contract does not support classified military or weapons-related work, but the agreement has still sparked internal protest. Debate over Israeli military AI use has intensified following the release of a new documentary *Naza*, which alleges that AI-powered targeting systems have contributed to high civilian casualties in the Gaza Strip. Israeli government officials have rejected the documentary’s claims as false and distorted.
As AI development accelerates, the internal backlash at Google highlights a growing rift between big tech’s public ethical commitments and the growing commercial demand for AI in national security and military contexts. With more tech firms pursuing military contracts, experts expect internal dissent over ethical AI use to continue growing across the industry.