Seven leading artificial intelligence companies have signed agreements with the Pentagon to integrate their technology into the US military’s classified networks, part of a broader push to transform the American armed forces into an “AI-first fighting force.” The deals, announced on Friday, involve SpaceX, OpenAI, Google, Nvidia, Reflection AI, Microsoft and Amazon Web Services.
Under the terms, each company has consented to the US military’s deployment of its technology for “any lawful use.” The Pentagon said in a statement that the agreements “accelerate the transformation toward establishing the United States military as an AI-first fighting force and will strengthen our warfighters’ ability to maintain decision superiority across all domains of warfare.” The technologies will be integrated into the Pentagon’s highest-tier classified environments — known as Impact Levels 6 and 7 — which handle secret intelligence analysis and sensitive national security data. According to a federal statement, the integration is intended to “streamline data synthesis, elevate situational understanding, and augment warfighter decision-making in complex operational environments.”
None of the participating companies are granted a veto over how the Pentagon ultimately uses their technology. The agreements include stipulations that AI should not be used for domestic mass surveillance or autonomous weapons without appropriate human oversight, but the precise deployment of each company’s tools has not been specified. The US Department of Defense is budgeting tens of billions of dollars for technology firms’ cutting-edge programs related to intelligence, drone warfare, and classified and unclassified information networks. It has requested $54 billion for the development of autonomous weapons alone.
A standoff over ethics and contract clauses
The most contentious element of the Pentagon’s AI push has been its dispute with the startup Anthropic, maker of the Claude chatbot. Anthropic rejected including the “lawful use” standard in its contract with the Defense Department, objecting over concerns that its technology might be used for domestic mass surveillance or fully autonomous lethal weapons. In response, the Pentagon designated Anthropic a supply-chain risk in February 2026 — the first time an American company has received such a designation. The move bars the Pentagon and its contractors from using Anthropic’s products, though they remain difficult to extricate from classified networks.
Anthropic sued the federal government, arguing the designation was retaliatory and amounted to an overreach. A federal court initially granted a preliminary injunction blocking the ban, but a court of appeals later denied Anthropic’s request to lift the designation. Officials at the Pentagon believe that signing agreements with Anthropic’s rivals could bring the holdout startup back to the negotiating table, according to the New York Times. The Pentagon’s actions against Anthropic have also spurred increased interest in other AI startups, with the military reportedly speeding up the process of incorporating new entrants onto secret and top-secret data levels — reducing the timeline from 18 months or longer to less than three months.
Anthropic’s latest AI model, the cybersecurity-focused Mythos, has rattled government officials and bankers over its ability to find vulnerabilities in well-tested software. The model’s release has complicated efforts by President Donald Trump and Defense Secretary Pete Hegseth to blacklist Anthropic. Separately, the UK government is reportedly seeking to deepen ties with Anthropic following its dispute with the US Pentagon, recognising its lack of a domestic AI lab on par with leading American companies.
New entrants and internal dissent
One of the seven companies, Reflection AI, has yet to release a publicly available model. Founded in 2024 by former Google DeepMind researchers, the two-year-old company aims to create open-source models as a counter to Chinese AI firms such as DeepSeek. It is seeking a $25 billion valuation, according to the Wall Street Journal, and has raised over $2 billion across multiple funding rounds. Nvidia led a $1 billion round with a $500 million investment, joined by 1789 Capital — the venture fund where Donald Trump Jr. is a partner — and DST Global. Reflection AI is also a founding member of Nvidia’s Nemotron Coalition.
SpaceX became a large-language-model provider earlier in 2026 after merging with xAI Holdings Corp., acquiring the Grok family of language models. It may also acquire Cursor’s Composer series of coding models. Google, whose Gemini model already powers the Pentagon’s internal AI portal GenAI.mil, signed the deal despite internal opposition: a group of Google employees sent an open letter to chief executive Sundar Pichai urging him to refuse classified military work, citing concerns about inhumane or harmful uses of AI. OpenAI had previously secured a deal to provide AI systems for classified Pentagon environments, which was later tweaked to prevent its tools from being used for domestic mass surveillance or autonomous weapons without appropriate human oversight.
A strategic framework and its critics
The agreements form a key component of the AI acceleration strategy unveiled by Defense Secretary Pete Hegseth in January. Hegseth said the strategy would “unleash experimentation, eliminate bureaucratic barriers, focus on investments, and demonstrate the execution approach needed to ensure we lead in military AI and that it grows more dominant into the future.” The strategy is built on three pillars — warfighting, intelligence, and enterprise operations — and is being executed through seven “Pace-Setting Projects.” The Pentagon’s internal AI portal, GenAI.mil, which runs on Google’s Gemini model, has already been used by more than 1.3 million Defense Department personnel since its launch, facilitating the creation of hundreds of thousands of AI agents.
The plans have sparked controversy and concerns over public spending, global cybersecurity and the potential for AI technology to be used for domestic surveillance. Critics have also raised alarms about the capacity for fully autonomous lethal weapons. The UK Ministry of Defence, meanwhile, has its own AI ethics principles — including human-centricity, responsibility, understanding, bias and harm mitigation, and reliability — and has awarded contracts to over two dozen companies for the development of AI targeting and decision-making capabilities under its Asgard programme. That programme has drawn criticism from groups such as Drone Wars UK over concerns about potential increases in civilian casualties caused by AI-driven targeting.
