Google is in discussions with the U.S. Department of Defense about integrating its Gemini system into Pentagon operations across all levels of information access, from open to highly classified. This was reported by The Information citing sources.
The agreement would allow the Pentagon to utilize Google's technology "for all lawful purposes."
During negotiations, Google proposed additional provisions to prevent the use of Gemini for mass surveillance of citizens or the development of autonomous weapons.
A Pentagon representative declined to comment directly on the talks with Google but emphasized that the department plans to continue implementing advanced AI technologies through close collaboration with the private sector at all levels of classification.
On April 28, 2026, it was reported that Google signed an agreement with the U.S. Department of Defense, allowing the Pentagon to use AI "for any lawful government purposes," according to The Information citing a source.
The corporation has joined OpenAI and xAI, which also provide their technologies for use in classified environments.
The agreement requires Google to assist in configuring security parameters and AI filters at the government's request. The contract includes a clause stating:
"The parties agree that the AI system is not intended and should not be used for mass surveillance within the country or autonomous weapons (including target selection) without proper human oversight and control."
However, the agreement "does not grant any rights to control or veto lawful operational decisions of the government."
Google proceeded with the deal despite employee opposition. On April 27, over 600 employees sent a letter to CEO Sundar Pichai demanding that the Pentagon be prohibited from using AI in classified environments.
Anthropic Controversy
A recent conflict between the Pentagon and AI startup Anthropic arose over ethical boundaries regarding the application of models. This escalated into a legal dispute and a ban by President Donald Trump on the use of the company's technologies across all federal agencies.
The roots of this issue date back to July 2025, when the U.S. Department of Defense awarded contracts worth up to $200 million to Anthropic, Google, OpenAI, and xAI for developing AI solutions in security. The department's Digital and AI Technology Office planned to use their solutions to create agent systems. Of all the contractors, only Anthropic's tools were integrated into classified environments due to their high quality.
However, by January 2026, the WSJ reported on the risk of contract termination. Disagreements arose due to the startup's strict ethical policies: Anthropic's rules prohibit the use of the Claude model for mass surveillance and in autonomous lethal operations.
Frustration among officials grew amid the integration of the Grok chatbot into the Pentagon's network. Defense Secretary Pete Hegseth emphasized that the department "will not use models that do not allow for warfare."
The situation intensified in February 2026 when the U.S. military employed Claude in an operation to capture Venezuelan President Nicolás Maduro.
CEO of the startup Dario Amodei stated that Anthropic would prefer not to collaborate with the Pentagon than agree to the use of its technologies in ways that could "undermine rather than protect democratic values." He confirmed that the issue lies in the potential use of tools like Claude for two purposes: "mass surveillance within the country" and "fully autonomous weapons."
Anthropic's principled stance has led to increased popularity of the startup's products, as users appreciate the developers' commitment to defending their interests.
Meanwhile, OpenAI faced criticism for its agreement with the Department of Defense, with Sam Altman later calling the decision hasty.
Return to Roots
The current negotiations between Google and the Pentagon echo events from 2018, when the corporation withdrew from Project Maven—a program that utilized artificial intelligence to analyze drone footage.
The company abandoned the project due to significant employee dissatisfaction. Over 3,000 individuals signed an open letter to management demanding an end to the development of AI for combat drones. This incident sparked a widespread discussion in the tech industry about the acceptability of using algorithms for military purposes.
Since then, Google has been gradually rebuilding its relationship with the government. In 2022, the company established a special division for working with the public sector.
In 2024, the firm signed a cooperation agreement in AI with the Department of Defense, which was limited to non-classified applications.
In early 2025, Google removed from its internal AI principles a clause that explicitly prohibited the use of technology in "weapons and surveillance systems."
The potential signing of a contract allowing work in classified environments indicates a move beyond typical "collaboration." This step suggests readiness for full-scale integration of the company's technologies into key defense operations.
Employee Opposition
Google's involvement in defense contracts continues to provoke internal dissent.
In 2018, the company promised never to work on military technologies. However, in April 2024, it was revealed that the company provided cloud computing services and access to AI tools to the Israeli Ministry of Defense.
In response, around 200 employees from DeepMind, Google's AI division, officially protested. They called for the termination of agreements with defense agencies, fearing that the lab's advanced developments were being handed over to armies engaged in real combat operations.
"Any involvement in military and weapons production negatively impacts our position as a leader in ethical and responsible AI, and contradicts our mission and stated principles of artificial intelligence," reads the document, which gathered signatures from about 5% of employees.
Pentagon Insists
In February, media reported that the Pentagon is urging leading AI companies to make their tools available in classified environments without the standard restrictions typically applied to users.
Such networks are used for a wide range of confidential tasks, such as operational planning or weapon targeting. The military wants to leverage AI's ability to synthesize information to aid decision-making.
However, neural networks can make mistakes and generate false information. The use of such tools in combat situations carries fatal risks.
AI companies strive to minimize the downsides of their products, but the Pentagon is frustrated with such limitations.
It is worth noting that the Department of Defense intends to accelerate decision-making through artificial intelligence and make it a central element of its strategy to integrate sensors and strike systems in combat operations.