Google DeepMind staff quit over military AI work: 'Ashamed to work at Google'

Internal opposition to Google DeepMind's military AI work is contributing to departures from the company, the BBC reported.

"Realizing my company is pushing more and more for AI systems in military operations makes me feel ashamed to work at Google," a London-based research engineer told the broadcaster. 

Here's what to know

In 2025, Google entered into a Pentagon agreement that permits its AI models to be used in classified military environments, a move that triggered employee criticism. 

Under the deal, the tools may be used for "any lawful government purpose," but not for domestic mass surveillance or autonomous weapons, including target selection, unless there is "appropriate human oversight and control," the BBC reported.

After three years at Google DeepMind, London-based research engineer Sofia Liguori told the BBC she quit because the firm's work with the U.S. and Israeli militaries was increasing.

"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," she admitted.

Another former Google DeepMind AI researcher, Andreas Kirsch, expressed similar concerns after resigning, saying he believes "Google's governance and oversight are not sufficient or meaningful enough to deploy AI safely and responsibly in military and classified settings," per the BBC.

Companies that signed similar military AI agreements in 2025 included Reflection, Nvidia, Oracle, SpaceX, Microsoft, Amazon, and OpenAI.

More background

A central concern is how quickly AI is moving into military systems.

One example is Maven Smart Systems. Palantir developed the Pentagon platform to organize intelligence and data and to help personnel identify and select strike targets using AI. 

Retired Lieutenant General Jack Shanahan, the former director of Project Maven, described what he saw as the "philosophy" behind the efforts.

"Go as fast as possible to deploy [AI technology] as widely as possible, potentially at the expense of the rigor and discipline of test and evaluation," Shanahan explained, per the BBC.

Sarah Shoker, formerly head of OpenAI's Geopolitics team, warned that AI can compress decision-making time in warfare.

"When you have hundreds of targets that are selected by AI and a short amount of time, it is almost impossible for a human operator to meaningfully review the AI's outputs for errors," she said, according to the BBC.

AI systems are increasingly intertwined with physical infrastructure, especially the energy grid. Running and scaling AI requires massive computing power and water-intensive cooling, straining utilities and other public resources.

Supporters argue AI can help optimize complex systems, including cleaner energy networks. The technology may improve planning and efficiency in some settings while also creating security risks, misuse concerns, and potentially driving up energy costs as demand grows.

In 2024, Google employees also criticized the company over Project Nimbus, a $1.2 billion joint cloud and AI contract with Amazon for Israel's government and military.

What's being done?

Google claims it has formal guardrails in place. 

A company spokesperson told the BBC that it uses "a consistent process" to review public sector contracts and evaluates them under its acceptable use policies and AI principles.

Google's principles suggest its AI should include "human oversight" and align with "international law and human rights." 

The Pentagon contract explicitly holds that the company's models "should not be used for domestic mass surveillance or autonomous weapons (including target selection) without appropriate human oversight and control," the BBC reported.

The company also said employees can use "dedicated channels" to raise ethics or misconduct concerns, and that retaliation is prohibited under strict workplace policies.

Critics argue the real question is whether those safeguards stay meaningful once AI tools become embedded in classified military systems.

Pentagon leaders have made clear they want faster adoption. 

"Speed defines victory in the AI era," Emil Michael, a Department of Defense undersecretary, told the BBC earlier this year. 

Bloomberg and others reported that an "overreliance on AI" and lack of human thoroughness, prioritizing speed over accuracy, led to the bombing of a school in Iran that killed at least 123 children in February. The root cause was reportedly an outdated database identifying the school, built in 2017, as being a military site since the land was previously used by the Islamic Revolutionary Guard Corps in past decades.

"Current AI models are not robust enough, and I don't think they have been trained for such tasks either," Kirsch cautioned.

Where can I learn more?

These articles dig deeper into the fallout over military AI, employee pushback, and Google's broader ambitions. They cover resignations and protests across big tech, along with debates over Pentagon adoption.

• At Google, a security veteran quit, saying leaders lost their moral compass on Pentagon AI.

• At Amazon, more than 1,000 employees warned AI could fuel a more militarized surveillance state.

• The Pentagon moved to adopt Elon Musk's Grok, extending AI use in military operations despite backlash.

Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.