LONDON, United Kingdom — Researchers at Google DeepMind are resigning over the company’s expanding military artificial intelligence contracts, citing concerns about the use of its technology in warfare.
Several current and departing employees told the BBC in an investigation published on Wednesday, October 21, 2026, that they were troubled by Google’s agreements with the United States military and its work with Israel.
Among those leaving are Sofia Liguori, a London-based research engineer who resigned this week, and Andreas Kirsch, an AI researcher who departed earlier in October.
Both questioned whether Google had adequate safeguards to control how military organisations use its artificial intelligence systems.
Google defended its contracts, saying they undergo reviews for compliance with its ethical policies and that employees have established channels for reporting concerns.

Researchers Question Military Use of AI
Liguori, who worked at DeepMind for three years, said her decision followed growing unease about Google’s relationships with American and Israeli defence organisations.
She told the BBC that the company’s increasing involvement in military applications of artificial intelligence made her ashamed to remain an employee.
The researcher said she had lost confidence in Google’s commitment to its publicly stated ethical principles.
Kirsch raised separate concerns about the company’s ability to monitor the deployment of its technology in classified military environments.
He questioned whether Google’s governance arrangements were sufficiently robust to prevent inappropriate applications of its AI systems.
Kirsch also said the company appeared to have limited information about how the Pentagon deployed its technology and lacked clear authority to prevent particular uses.
He argued that existing AI models had not been adequately tested for battlefield operations.
Their departures follow a broader disagreement among employees about Google’s military partnerships and the company’s responsibilities as a developer of advanced artificial intelligence.

Pentagon Agreement Permits Classified AI Operations
Google signed an agreement with the US Department of Defense earlier this year allowing its AI technology to be deployed in classified military environments.
The contract permits use for lawful government purposes while establishing conditions concerning domestic mass surveillance and autonomous weapons, including requirements for appropriate human oversight and control.
The agreement is among several signed by technology companies seeking to supply artificial intelligence capabilities to the Pentagon.
Other companies involved in military AI agreements include OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia, and Reflection.
US Defence Secretary Pete Hegseth has made artificial intelligence a priority for the armed forces, calling for faster adoption of the technology across military operations.
The Pentagon has also expanded its use of AI systems to analyse intelligence and support operational decisions.
One such system is Maven Smart Systems, developed by Palantir, which uses artificial intelligence to process intelligence and assist military personnel in identifying potential targets.
The Pentagon maintains that rapid adoption of the technology is necessary for military effectiveness.
But some researchers and former defence officials have questioned whether the speed of deployment allows sufficient time to evaluate the reliability of these systems.
Sarah Shoker, a former head of OpenAI’s geopolitics team, told the BBC that automated systems could dramatically reduce the time available for human operators to review targeting recommendations.
She warned that when AI produces large numbers of potential targets in rapid succession, military personnel may struggle to identify errors before decisions are made.

Google’s Israeli Contract Faces Renewed Scrutiny
Google’s agreements involving Israel have also prompted opposition among employees.
The company and Amazon operate Project Nimbus, a cloud computing and artificial intelligence contract valued at $1.2 billion that provides technology to Israeli government and military organisations.
The arrangement attracted employee criticism in 2024 over concerns about potential military applications.
Google previously maintained that the contract was not intended for classified operations involving weapons or intelligence services.
The issue has received renewed attention amid examinations of Israel’s use of AI-supported targeting systems during military operations in Gaza.
A documentary titled Naza examined allegations concerning the use of such systems in attacks that caused civilian casualties.
An Israeli government minister disputed the documentary’s findings, describing its claims as false and distorted.
The BBC’s investigation did not establish that Google DeepMind’s technology had been improperly used in a particular military operation.

Employees Express Fears About Speaking Publicly
The BBC also interviewed current Google DeepMind employees who requested anonymity because they feared professional consequences.
Some said they were reluctant to criticise the company’s military work publicly because of restrictions in their employment agreements.
Liguori said her contract limited her ability to speak critically about Google, but she nevertheless decided to discuss her concerns.
One serving employee said researchers who had joined DeepMind because of its original scientific mission might leave as its direction changed.
Another told the broadcaster that colleagues dissatisfied with the company’s ethical decisions had already departed.
Neither employee provided a verified figure for the number of resignations.
Google rejected suggestions that workers lacked safe avenues for raising concerns.
A spokesperson said the company reviewed government contracts against its acceptable-use policies and AI Principles.
Those principles include commitments to human supervision, risk assessment, and respect for international law and human rights.
The company also said employees could report ethical concerns and misconduct through dedicated internal channels, and that workplace rules prohibited retaliation.
The BBC interviews document continuing disagreement within Google DeepMind over the company’s military activities, with departing researchers questioning whether its oversight mechanisms can adequately govern the use of AI in classified operations.





