Key Points
- Google DeepMind employees Sofia Liguori and Andreas Kirsch have resigned over concerns about Google’s growing involvement in artificial intelligence contracts with the US and Israeli militaries, according to BBC reporting by Christopher Giles, with additional reporting by Rosa Rahimi.
- London-based research engineer Sofia Liguori said she felt ashamed to work at Google because of the company’s increasing support for AI applications in military operations.
- AI researcher Andreas Kirsch questioned whether Google’s governance and oversight arrangements were sufficient to ensure responsible use of its technology in classified military environments.
- Google said it follows established procedures for reviewing public-sector contracts, complies with its acceptable-use policies and AI principles, and provides channels for employees to raise concerns.
- Google is among several technology companies that have agreed to supply AI capabilities for US military applications, alongside OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia and start-up Reflection.
- Concerns centre on the reliability of AI systems in warfare, the meaningfulness of human oversight, the potential for targeting errors and the limited visibility technology providers may have into how their systems are deployed.
- Google’s joint $1.2bn (£960m) Project Nimbus agreement with Amazon, which provides cloud computing and AI technology to the Israeli government and military, has also attracted employee criticism.
- The dispute reflects a wider debate across the technology industry about whether commercial AI development can be reconciled with military requirements, national security priorities and commitments to responsible AI.
- Earlier protests against military AI contracts, including Google’s Project Maven controversy in 2018, demonstrate that disagreements over the use of commercial technology in warfare have a history within the company.
Why Have Google DeepMind Employees Resigned Over Military AI Contracts?
Google is facing renewed internal opposition to its military artificial intelligence work after employees at Google DeepMind resigned over concerns that the company’s technology could be used in military operations without adequate safeguards. The dispute centres on the ethical responsibilities of AI developers, the reliability of AI systems in warfare and the extent to which technology companies can control how their products are used by governments.
As reported by Christopher Giles of BBC News, with additional reporting by Rosa Rahimi, London-based research engineer Sofia Liguori resigned from Google DeepMind during the week of 19 October 2026, while AI researcher Andreas Kirsch had left the company earlier in the month. Both raised concerns about Google’s expanding military commitments and whether its oversight arrangements were sufficient to prevent unsafe applications.
Liguori told the BBC that the company’s growing involvement in military AI had changed how she viewed her employment. “Realising my company is pushing more and more for AI systems in military operations makes me feel ashamed to work at Google,” she said, speaking in a personal capacity.
According to the BBC report published on 21 October 2026, Liguori had worked at Google for three years before deciding to resign. She cited the company’s increasing work with the US and Israeli militaries as a reason for her decision.
“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 told the BBC.
Her departure adds to continuing questions about the balance between commercial technology development, government contracts and employee expectations concerning the responsible use of artificial intelligence. The resignations do not, by themselves, establish that Google’s technology has been used improperly in a particular military operation. They do, however, highlight concerns raised by people who have worked within the company about its policies and the practical safeguards surrounding military deployments.
The issue is particularly significant for Google DeepMind, Google’s AI research business, as its technologies become increasingly relevant to applications extending beyond conventional consumer products and commercial software.
What Did Google Say About Its Military AI Contracts?
Google defended its approach to government contracting in a statement provided to the BBC, emphasising its review procedures, internal policies and mechanisms for employee concerns.
A Google spokesperson told the broadcaster: “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 spokesperson added that the company had dedicated channels for employees to raise concerns, including matters involving business ethics and misconduct. Google also said it had strict workplace policies intended to prevent retaliation against employees who raise such issues.
The response places the company’s position in contrast with the concerns expressed by Liguori and Kirsch. While Google maintains that its contracting processes and AI principles provide a framework for responsible work, the departing researchers questioned whether those arrangements offered sufficient practical control over military applications.
The BBC report did not establish that Google had breached its internal policies. Instead, it documented a disagreement about the adequacy of the company’s safeguards and its ability to ensure that AI systems are used consistently with its stated principles.
That distinction matters because the dispute concerns not only the ethical intentions of a technology provider but also the practical limits of oversight once AI systems are supplied to government customers.
For businesses developing advanced technologies, the issue also raises broader questions about how contracts should define permitted uses, what information suppliers should receive about deployment, and how concerns can be investigated when systems operate in classified environments.
Why Is Andreas Kirsch Concerned About Google’s AI Governance?
Andreas Kirsch, an AI researcher who resigned from Google DeepMind earlier in October, questioned whether the company had adequate governance arrangements for the deployment of AI in sensitive military settings.
According to Christopher Giles’s BBC report, Kirsch said he believed that “Google’s governance and oversight are not sufficient or meaningful enough to deploy AI safely and responsibly in military and classified settings” or “to develop artificial superintelligence safely in the future.”
Kirsch also expressed concern about the company’s ability to control how the US Department of Defence uses its technology.
“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,” he told the BBC.
His comments point to a structural challenge facing AI suppliers. A company may establish restrictions governing how its technology should be used, but enforcing those restrictions can become more difficult when customers operate in confidential or classified environments.
The concern is not limited to whether an AI system performs as intended in testing. It also involves how the system is integrated into operational procedures, who reviews its recommendations, whether errors can be detected and what happens when a customer uses the technology in circumstances that differ from those anticipated by its developers.
For a company such as Google, these questions intersect with its public commitments to responsible AI and its commercial relationships with governments.
The dispute also raises questions about the future development of increasingly capable AI systems. Kirsch linked his concerns about military deployment with the broader challenge of developing artificial superintelligence safely, indicating that he views governance as an issue extending beyond individual defence contracts.
However, his assessment represents his own position as a departing researcher. Google has defended its review procedures and said that its policies and employee reporting channels are designed to address concerns.
How Could AI Systems Affect Military Targeting and Battlefield Decisions?
The debate extends beyond Google’s internal policies to the technical limitations and operational risks associated with using AI in warfare.
As reported by Christopher Giles of BBC News, one prominent example of military AI infrastructure is Maven Smart Systems, the Pentagon’s principal software platform for organising intelligence and other data. Developed by US defence company Palantir, the system uses AI to help military personnel select and identify targets for strikes.
AI systems can process large volumes of information more quickly than human analysts, potentially helping military organisations organise intelligence, assess information and support operational planning. However, speed does not necessarily guarantee accuracy, and the consequences of an error can be particularly serious when a system contributes to decisions involving the use of force.
Sarah Shoker, formerly head of OpenAI’s Geopolitics team, warned the BBC about the implications of accelerating decision-making through AI.
“AI compresses the time an operator has to make a decision because it works much faster than a person, and that’s especially risky in a high-stress environment,” she said.
Shoker also highlighted the challenge of reviewing large numbers of AI-generated recommendations within a limited period.
“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 told the broadcaster.
Her comments highlight a central question in the debate over human oversight: whether a person can genuinely evaluate an AI recommendation when operational demands leave little time to examine the underlying information.
Human involvement in a decision-making process does not automatically mean that the oversight is effective. Its value depends on whether personnel have sufficient time, relevant expertise, access to supporting evidence and the authority to reject a system’s recommendation.
In military settings, the distinction can have significant consequences. An inaccurate recommendation might arise from incomplete information, flawed interpretation or weaknesses in the underlying model. If personnel rely too heavily on automated outputs, errors may be harder to identify before decisions are made.
Kirsch also argued that existing AI models were not sufficiently robust for battlefield use.
“Current AI models are not robust enough, and I don’t think they have been trained for such tasks either,” he told the BBC.
These are concerns raised by individuals interviewed for the report rather than a finding that every military AI system is unreliable. Nevertheless, they demonstrate why model performance, testing, human supervision and accountability have become central to discussions about the deployment of AI in defence.
The questions are also relevant to organisations operating in other high-risk environments, where decisions must be explainable, errors need to be identified quickly and responsibility cannot be transferred entirely to an automated system. The development of appropriate controls is therefore an important consideration in Artificial Intelligence, particularly where AI systems support consequential operational decisions.
Which Technology Companies Have Signed Military AI Agreements?
Google’s involvement forms part of a wider shift in the relationship between major technology companies and the US military.
According to Christopher Giles of BBC News, Google is among several companies that signed agreements with the US military earlier in 2026 to deploy AI across classified military environments. The companies named in the report include OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia and the start-up Reflection.
These agreements cover a range of military applications, including transport, planning and battlefield operations. The contracts illustrate how AI is becoming part of a broader technology infrastructure supporting defence organisations, rather than remaining confined to consumer services and conventional business software.
The Pentagon’s interest in AI has also been reinforced by political and military leadership. US Defence Secretary Pete Hegseth has made integrating AI into the armed forces a priority.
As reported by the BBC, Hegseth said in January that the Trump administration would become an “AI-first warfighting force”.
The policy direction creates commercial opportunities for technology providers seeking government contracts, while raising questions about how their products will be deployed and what restrictions should apply.
A separate report published by The Financial Times in April 2026 documented employee opposition to Google’s renewed military AI relationship. According to the newspaper, more than 560 employees signed an open letter to chief executive Sundar Pichai opposing the deal and expressing concerns about potentially inhumane applications of the technology. The FT reported that Google’s agreement extended an existing contract worth $200 million, allowing its AI to be used in classified operations.
The FT also reported that Alphabet president of global affairs Kent Walker had emphasised the need for responsible work and proper oversight. The company said its technology should not be used for domestic mass surveillance or autonomous weapons without human control.
Taken together, the reports show a dispute that extends beyond individual resignations. Employees have questioned the company’s involvement in classified work, while management has defended the commercial relationship and pointed to restrictions governing the use of its technology.
The wider industry faces a similar challenge: how to supply increasingly capable systems to government customers while maintaining meaningful safeguards and addressing concerns raised by staff, researchers and the public.
How Does Google’s Project Nimbus Agreement With Israel Fit Into the Controversy?
Google’s military AI debate also involves its relationship with the Israeli government.
According to Christopher Giles’s BBC report, Google’s joint agreement with Amazon, known as Project Nimbus, is worth $1.2bn (£960m) and provides the Israeli government and military with AI technology and cloud computing infrastructure.
The agreement has faced criticism from Google employees since 2024. At the time, the company said Project Nimbus was “not directed at highly sensitive, classified, or military workloads relevant to weapons or intelligence services”.
That statement is important when distinguishing the company’s stated position at the time from the broader concerns raised by employees about the agreement and its potential applications.
The BBC report also discussed the documentary Naza, which examines the alleged role of AI-powered targeting systems in bombardments in Gaza and the resulting high civilian casualties. An Israeli government minister said the claims presented in the film were false and distorted.
The documentary’s claims and the minister’s response represent competing positions in a wider debate over AI-assisted warfare. They should not be treated as proof that Google’s technology was responsible for any particular military action.
For Liguori, however, Google’s work with both the US and Israeli militaries contributed to her decision to leave the company. Her resignation reflects an ethical objection to the direction of its business relationships, rather than a finding of legal wrongdoing.
The controversy illustrates the difficulty of separating the provision of general-purpose technology from its eventual use by customers. Cloud computing and AI tools can support many different activities, but their applications in defence and intelligence can raise additional questions about transparency, oversight and accountability.
For technology companies, these issues may affect employee retention, public trust and the scrutiny applied to government contracts. They also demonstrate why the terms of an agreement, the safeguards attached to it and the ability to monitor deployment can be as important as the technology itself.
Are Google Employees Afraid to Speak Out About Military AI?
The BBC also reported concerns among current and former employees about the risks of publicly criticising Google’s military work.
Liguori said her employment contract restricted her ability to criticise the company. She told the broadcaster: “In principle I’m not allowed to criticise Google for the duration of my life.”
She said she had decided to speak publicly because she believed the issues should be known, adding that discussing facts was not the same as criticising the company.
Liguori also described an atmosphere in which employees could feel reluctant to raise their concerns.
“People are scared of retaliation and putting their career at risk. You feel you are being watched, you can’t speak,” she told the BBC.
These comments describe Liguori’s experience and perceptions. They do not independently establish that Google retaliated against employees who raised concerns.
Google, for its part, told the BBC that it had dedicated channels for employees to raise concerns about business ethics and misconduct and that strict workplace policies were in place to prevent retaliation.
The broadcaster also interviewed two current Google DeepMind employees who requested anonymity because they feared retaliation.
One employee warned that workers motivated by the company’s original mission might leave, while new recruits at the beginning of their careers could fill the resulting vacancies. The employee described the possibility of “a changing of the guard”.
Another current employee said that many people who felt strongly about changes to Google DeepMind’s ethical position had already left.
“Some people are putting up a fight, others are thinking: ‘it’s a big company. What do you expect to change?’”, the employee told the BBC.
The comments suggest that the disagreement could have implications for staff morale and recruitment, particularly among researchers who joined the company because of its ambitions for AI research and its stated commitment to responsible development.
Whether these concerns will lead to further departures remains uncertain. The BBC report did not establish how many additional employees intend to resign or quantify the impact on Google’s research operations.
Nevertheless, the issue raises questions for technology businesses about how to handle internal dissent, protect employees who raise good-faith concerns and maintain confidence in corporate governance when commercial decisions are contested.
What Does the History of Google’s Military AI Work Reveal?
The current dispute follows an earlier confrontation over Google’s military relationships.
In 2018, employee protests over Project Maven, a Pentagon initiative that used AI to analyse drone surveillance footage, helped persuade Google not to renew its involvement in the project.
A Fortune report by Beatrice Nolan, published on 4 May 2026, revisited that episode while examining the response to Google’s newer Pentagon agreement. Nolan reported that hundreds of employees had opposed the more recent arrangement, but that the company’s response differed from the earlier retreat.
The historical comparison illustrates how the relationship between major technology companies and defence organisations has evolved. As AI has become more commercially significant and governments have sought greater access to advanced systems, the incentives surrounding military contracts have also changed.
The dispute has also been affected by changes to Google’s stated AI commitments. In its coverage of an earlier employee protest, The Guardian reported in May 2026 that a former Google DeepMind engineer had alleged unfair dismissal after opposing the company’s work for Israel. The engineer’s account was disputed by Google, which said it did not accurately reflect the facts.
That separate employment dispute demonstrates that concerns about military AI have extended beyond resignations to questions about workplace rights and how companies respond to internal criticism. It should, however, be distinguished from the resignations described in the BBC’s October report.
The recurring disagreements show that the issue is not simply whether AI can be used for defence purposes. It also concerns which applications should be permitted, what restrictions should apply, how those restrictions are enforced and how employees can challenge decisions they consider inconsistent with a company’s principles.
Google’s present position is that it reviews government contracts and applies its acceptable-use policies and AI principles. The departing researchers argue that those arrangements do not provide sufficient assurance in classified military settings.
The two positions reflect a broader challenge facing technology companies as their products become increasingly relevant to national security.
What Could the Resignations Mean for the Future of Military AI?
The immediate impact of Liguori’s and Kirsch’s departures on Google’s operations is not clear from the BBC report. There is insufficient information to determine whether their resignations will affect specific projects, delay contracts or change the company’s military AI strategy.
The wider implications, however, extend to technology governance, employee confidence and the relationship between commercial AI developers and government customers.
First, the dispute highlights the importance of meaningful human oversight. A policy requiring human involvement is unlikely to resolve every concern if personnel cannot adequately assess a system’s recommendations or intervene when errors occur.
Secondly, the controversy raises questions about contractual accountability. Technology providers may need to consider how permitted uses are defined, what monitoring arrangements are feasible and how concerns are handled when systems operate in classified environments.
Thirdly, employee trust may become an important factor in the development of advanced AI. Researchers who disagree with a company’s direction may leave, while others may choose to remain and attempt to influence internal policy. The extent to which these disagreements affect recruitment and retention will depend on future decisions and employee responses.
Finally, the debate underscores the difference between a technology’s capabilities and the conditions in which it is deployed. An AI model that performs well in a particular test may still require extensive evaluation before being used in a complex, high-pressure environment.
For companies developing or implementing AI, these considerations reinforce the importance of governance frameworks, risk assessment, testing, transparency where possible and clearly defined responsibilities. They also explain why organisations working with advanced technology must consider not only what their systems can do, but how they are controlled and who remains accountable for their use.
Google’s military contracts remain part of its broader business activities, and the company has defended its approach to public-sector work. The resignations reported by the BBC nevertheless demonstrate that questions about the responsible deployment of AI remain contested within the technology industry.
The central issue is whether safeguards and oversight can keep pace with the growing use of AI in military operations. As governments expand their use of the technology, technology providers, military organisations and policymakers will continue to face questions about reliability, accountability and the limits of automation in decisions involving human lives.