In a major reversal of its previous stance on military work, Google is negotiating a deal to provide its most advanced AI to the Pentagon for use in classified environments.
Back
Back
Google in talks to deploy Gemini AI for US military work
(P1) Google is in advanced talks with the US Department of Defense to deploy its Gemini AI model in classified settings, a landmark shift for the tech giant that could significantly expand its role as a defense contractor and intensify its competition with Microsoft and OpenAI.
(P2) The negotiations are progressing positively and would permit the Pentagon to use Google's AI for "all lawful purposes," according to people familiar with the matter who spoke to The Information.
(P3) The proposed contract language mirrors terms secured by rival OpenAI, which recently signed its own deal with the Pentagon. The move contrasts sharply with AI safety-focused Anthropic, which was recently designated a "supply chain risk" by the Defense Department after refusing to remove contractual restrictions on the use of its AI in weapons systems.
(P4) A final agreement would bolster Google's public sector business, a division aiming for $6 billion in new contracts between 2025 and 2027. While a fraction of parent company Alphabet's projected $403 billion in 2025 revenue, the deal represents a critical effort to gain ground in the government cloud market, where Google's 14% share trails Amazon's 28% and Microsoft's 21%.
From Maven Protests to Renewed Military Partnership
Google's relationship with the US military has undergone a dramatic reversal since 2018, when the company dropped the Pentagon's Project Maven contract after intense employee protests over using AI to analyze drone footage. The decision created deep distrust within the Defense Department and sparked a wave of employee activism across the tech sector.
In the years since, Google has methodically rebuilt its defense ties. It established a dedicated public sector division in 2022, led by 30-year defense industry veteran Karen Dahut, and staffed it with former government contractors and military veterans. The division reportedly fosters a distinct internal culture, nicknamed "Big Google" by its members, to shield its work from internal dissent and limit access for non-US employees. This new negotiation follows a non-classified AI deal signed last November and a quiet revision of Google's AI principles in early 2025 that removed language explicitly forbidding AI use in weaponry.
Competitive Landscape and Internal Hurdles
The proposed deal's terms are said to be nearly identical to those OpenAI secured, reflecting a push by CEO Sam Altman for standardized contracts across the AI industry. However, the practical limitations of these "safety" clauses remain contentious. Lawyers have noted that language prohibiting fully autonomous weapons is not necessarily binding when the overarching agreement allows for "all lawful purposes," leaving a gray area that Anthropic refused to enter.
Despite the strategic push, Google's public sector unit remains a minor part of its overall business and faces significant headwinds. The division recently conducted layoffs after its growth failed to meet the broader cloud unit's targets. Former employees note that federal agencies are deeply entrenched with Amazon Web Services and Microsoft products, and that Google does not always match the on-site staffing levels of its competitors to facilitate government projects. To support the potential Pentagon work, Google may need to expand its limited classified infrastructure, potentially adding more GPUs and deploying its custom Tensor Processing Units (TPUs) in a secure environment for the first time.
This article is for informational purposes only and does not constitute investment advice.
