Google DeepMind Researchers Quit Over Military AI Contracts
DeepMind scientists are resigning over Google's military AI contracts after the company quietly dropped its no-weapons pledge.
Policy · Source: MadRobot (reporting BBC News investigation)
What happened
Google is bleeding top AI talent over its aggressive push into military contracts. DeepMind research engineer Sofia Liguori and AI researcher Andreas Kirsch both quit this month. They explicitly cited the company's growing involvement with the US and Israeli militaries as their reason for leaving. Liguori spent three years at the company before resigning. She stated that the shift toward military operations made her ashamed to work at Google. Kirsch added that Google lacks the oversight to deploy AI safely in classified settings. He warned that the company has no real control over how the Pentagon uses its models.
The resignations follow a major internal policy shift that went largely unnoticed. In February 2025, Google completely rewrote its AI Principles. The company quietly deleted a crucial 2018 clause that banned the development of AI for weapons and mass surveillance. Now, Google allows its models to be used for any lawful government purpose. This opens the door for sweeping defense contracts. Other major players like OpenAI, Amazon, Microsoft, Nvidia, Oracle, and the startup Reflection have signed similar agreements. The Pentagon is actively setting up commands for autonomous warfare and shopping for AI agents that can hack independently.
Employees are speaking out despite facing strict corporate gag orders. Liguori told the BBC her contract forbids her from criticizing Google for the duration of her life. She chose to speak anyway because she believes the public needs to know the facts. Anonymous staff report a deep culture of fear and retaliation inside DeepMind. They warn that many more principled engineers are preparing to leave. One employee noted that while motivated staff are walking out, a new pipeline of early-career engineers will likely replace them. This signals a complete changing of the guard at Google DeepMind.
Key facts
- February 2025 — When Google dropped its ban on AI for weapons and surveillance.
- $1.2bn — Value of the Project Nimbus cloud deal Google and Amazon share with Israel.
- 2018 — Year Google employees originally protested Project Maven and forced the first AI principles.
Why it matters
The talent war in artificial intelligence is colliding directly with the actual business of war. Top engineers usually want to build consumer products and scientific breakthroughs. They do not want to build targeting systems or autonomous hacking agents. When tech giants pivot heavily to defense contracts, they risk losing the very researchers who build their core assets. AI compresses decision times in high-stress environments. Experts warn that human operators cannot meaningfully review AI outputs when hundreds of targets are selected in seconds. This ethical friction creates a massive opening for startups. Founders who maintain clear ethical boundaries have a unique recruiting advantage to poach disillusioned top-tier talent from big tech.
The guardrails are officially off for enterprise AI in the defense sector. The Pentagon declared its intent to become an AI-first warfighting force. This normalizes military AI deployment across the entire industry. Defense tech is no longer a niche sector. It is rapidly becoming the primary revenue driver for foundational models. In 2018, employee protests forced Google to drop Project Maven and establish strict AI principles. Today, the corporate stance has hardened. Companies are prioritizing lucrative government deals over employee dissent. If you are building AI infrastructure today, the end user is increasingly likely to be a military operator.
For builders
Poach disillusioned big tech talent
Google and other giants are losing principled engineers over defense contracts. Startups with clear ethical boundaries have a unique recruiting advantage. You can attract top-tier researchers who refuse to build military applications. Make your company values explicit to win this talent.
Defense tech is a massive market
The Pentagon is actively shopping for autonomous AI agents and battlefield operations tech. If you are willing to navigate government contracts, the available budget is massive. Startups like Reflection are already signing deals alongside major players like OpenAI and Microsoft.
Expect strict lifetime gag orders
Big tech companies are enforcing aggressive lifetime non-disparagement clauses on their engineers. If you partner with or get acquired by these giants, read the fine print carefully. Your team might lose their right to speak publicly about their work forever.
My take
I see Google caving to the money. In 2018, employee protests actually stopped military contracts. Now, I watch them delete their own ethics rules and tell engineers to stay quiet for life. If you build foundational AI today, you are building defense tech whether you like it or not.
Original reporting: MadRobot (reporting BBC News investigation). This is my rewrite and opinion.