Google AI Researchers Are Quitting Over Military Deals

Google AI Researchers Are Quitting Over Military Deals

What happens when the people building artificial intelligence start worrying about how their own technology could be used?

That question is creating tension inside Google DeepMind, Google's artificial intelligence research division, after several employees resigned or began preparing to leave over the company's growing involvement in military AI projects.

In a report published on October 21, 2026, the BBC spoke to current and former workers who raised concerns about Google's contracts with the US military, the potential use of AI in warfare, and whether existing safeguards are strong enough.

Among them was Sofia Liguori, a London-based research engineer who resigned after three years at Google DeepMind. She told the BBC that the company's growing involvement in military AI made her ashamed to work there.

Another researcher, Andreas Kirsch, also left with concerns about whether Google's oversight systems could ensure that its AI technology was used safely in classified military settings. Their departures highlight a difficult question facing the technology industry: Just because AI can help a military operate faster, does that mean it is ready to make decisions in situations where human lives are at stake?

Why Are Google Employees Walking Away?

For some researchers, the concern is not simply that Google works with governments. It is how AI could be used once it enters military systems.

AI can process large amounts of information, identify patterns, and help people make decisions more quickly. In a military setting, these capabilities could support logistics, intelligence analysis, operational planning, and other activities. But problems arise when AI-generated information influences decisions involving potential targets or the use of force.

AI systems can make mistakes. They can misinterpret information, produce unreliable conclusions, or fail when circumstances differ from the situations they were trained to handle.

In an ordinary task, an error might mean correcting a document or repeating an analysis. In warfare, an error could have far more serious consequences. Liguori said Google's expanding military work had weakened her confidence in the company's commitment to its stated ethical values.

Kirsch raised concerns about the strength of Google's internal oversight and whether the company could meaningfully control how its technology was used after it entered classified military environments.

Their views do not represent every Google employee, but they reveal a growing disagreement over the responsibilities of companies developing powerful AI systems.

What Does Google's Military Contract Allow?

Google is among several major technology companies that signed agreements with the US military in 2026 to provide AI technology for classified work.

The US Department of Defense has made AI adoption a priority, arguing that the technology can help its forces analyze information and respond more quickly.

Under the reported Google agreement, its AI models can be used for “any lawful government purpose.” The contract also states that the technology should not be used for domestic mass surveillance or autonomous weapons, including target selection, without appropriate human oversight and control.

Google says its public-sector contracts go through a review process and must comply with its acceptable-use policies and AI principles. The company also says employees have channels for raising ethical concerns and that workplace policies prohibit retaliation.

However, critics question whether these protections are sufficient when AI is deployed in classified environments. The concern is that companies may have limited visibility into how their technology is used after it is integrated into military systems. If a developer cannot easily monitor a system's use or intervene in its operation, enforcing restrictions may become more difficult.

That is the heart of the disagreement: Google says it has policies and safeguards, while some employees believe those protections may not provide enough control over high-risk military applications.

How AI Is Already Being Used in Warfare

The debate is not purely theoretical. AI-enabled systems are already being used to support military operations.

One example is Maven Smart System, a platform developed by defense technology company Palantir. It helps military personnel organize and analyze intelligence, including information that can support the identification and assessment of potential targets.

AI can help process large volumes of information much faster than a person working alone. That speed can be useful when military personnel need to make sense of complex and rapidly changing situations. But speed can also create pressure.

If a system produces recommendations faster than people can carefully review them, human oversight risks becoming a quick approval process rather than a meaningful check.

Sarah Shoker, formerly head of OpenAI's Geopolitics team, warned the BBC that AI can compress the time available for an operator to make a decision. When many potential targets are being assessed quickly, she argued, it can become extremely difficult for a human to examine every output thoroughly.

Retired US Lieutenant General Jack Shanahan, who previously directed Project Maven, also expressed concern about the push to deploy AI widely and quickly, potentially at the expense of rigorous testing.

The issue is not that every AI-supported military system automatically makes decisions independently. Rather, it is whether the people responsible for those decisions have enough time, information and authority to challenge the technology when it gets something wrong.

Google's AI Ethics Are Under Scrutiny

Google's own AI principles emphasize human oversight and alignment with international law and human rights. Some employees believe the company's military partnerships raise questions about how those principles are applied in practice. This is not the first time Google's work with military and government customers has caused internal controversy.

In 2018, employee protests contributed to Google's decision not to renew Project Maven, an earlier Pentagon initiative involving AI-assisted analysis of drone footage.

More recently, Google and Amazon's Project Nimbus cloud-computing agreement with the Israeli government and public sector has faced criticism from employees concerned about the potential use of the technology in military operations.

Google has said Project Nimbus is not intended to support highly sensitive or classified military workloads relevant to weapons or intelligence services. Critics have questioned whether assurances about the limits of these contracts provide enough transparency.

The current dispute reflects a broader challenge for technology companies: publishing ethical principles is one thing; ensuring those principles remain meaningful when products are used in complex, high-stakes environments is another.