600+ Google employees protest Gemini AI use in Pentagon ops
The employees cite risks of misuse, lack of oversight, and potential harm to civilians.
-
External view of the Google building in Chelsea on Thursday, Oct. 09, 2025, in New York (AP)
More than 600 employees at Google have called on the company to abandon a potential agreement with the United States Department of War that would expand the use of its artificial intelligence tools into classified military operations, according to a statement released Monday.
The appeal, addressed to Chief Executive Sundar Pichai, was signed by staff across multiple units, including DeepMind and Cloud, and includes more than 20 individuals in senior leadership roles. It comes as the company is in discussions with US defense officials over deploying its Gemini AI system in sensitive, restricted environments.
Workers behind the initiative warned that the nature of classified programs makes meaningful oversight impossible. “Classified workloads are by definition opaque,” one organizing employee, who was not named in the statement, told AFP. “Right now, there's no way to ensure that our tools wouldn't be leveraged to cause terrible harms or erode civil liberties away from public scrutiny. We're talking about things like profiling individuals or targeting innocent civilians.”
Gemini deal backlash
The proposed deal would build on Google’s existing collaboration with the Pentagon under the genAI.mil program, which currently focuses on non-classified applications. Extending those capabilities into classified domains has raised concern among employees who argue that internal safeguards cannot be reliably enforced once the technology is handed over.
According to organizers, Google has suggested contract provisions intended to limit the use of Gemini for domestic mass surveillance and fully autonomous weapons systems without human oversight. However, US defense officials are said to favor broader language allowing “all lawful uses,” a formulation employees believe would weaken any meaningful restrictions.
Staff members also pointed to Pentagon policies that prevent external actors from imposing binding constraints on how its AI systems are used. “If leadership is truly serious about preventing downstream harms, they must reject classified workloads entirely for now,” a second letter organizer said.
AI warfare concerns
The pushback comes amid growing scrutiny over the use of AI-assisted systems in military operations, particularly after investigations into the Israeli occupation’s war on Gaza revealed how algorithmic tools were used to accelerate targeting decisions.
In 2024, +972 Magazine and Local Call reported that the Israeli military used an AI system, known as Lavender, to mark tens of thousands of Palestinians in Gaza as "suspected targets", with limited human review and a permissive approach toward civilian casualties. The Guardian also reported that Israeli intelligence sources described the system as part of a process in which AI-generated recommendations were used to support airstrikes in densely populated areas.
Another Israeli system, known as The Gospel, was reported to have helped generate bombing targets at a far greater pace than previous human-led methods, fueling concern that AI tools can turn target production into an industrialized process with devastating effects on civilians.
More recently, the deadly strike on the Shajareh Tayyebeh school in Minab, southern Iran, has renewed questions about AI-assisted targeting and accountability. AP reported that evidence suggested the blast, which killed more than 165 people, mostly children, was likely caused by a US airstrike. A US Senate letter also cited concerns over the reported use of AI tools to select and prioritize targets in Iran.
These cases have reinforced the concerns raised by Google employees that once AI systems are integrated into classified military environments, external scrutiny becomes nearly impossible and responsibility for civilian harm becomes harder to establish.
Defense race
The dispute unfolds as competition intensifies among major technology firms to secure defense contracts, particularly following tensions between the Pentagon and Anthropic, which has challenged its designation as a “supply-chain risk” after opposing certain military and surveillance applications of its AI systems.
The current employee push draws parallels with the 2018 backlash against Project Maven, when internal pressure led Google to withdraw from a Pentagon initiative involving AI-assisted drone analysis. In the years since, the company has gradually expanded its defense-related activities, positioning itself alongside rivals such as Microsoft and Amazon Web Services in the race for government AI contracts.
Read more: Google plans up to $40bn investment in Anthropic AI