The Pentagon said Friday that it has reached deals with seven technology companies to bring artificial intelligence tools into U.S. military classified computer networks, with the Defense Department framing the change as a way to improve how service members make decisions in complex battlefield conditions.
In the statement, the Defense Department said the companies will provide resources intended to “augment warfighter decision-making in complex operational environments.” The Pentagon named Google, Microsoft, Amazon Web Services, Nvidia, OpenAI, Reflection and SpaceX as part of the effort.
The Pentagon’s announcement also highlighted who is not in the mix. Anthropic was notably absent from the list after, as described in the report, a public dispute and a legal fight between the company and the Trump administration over how the U.S. government should use AI in war and related safeguards.
The move lands amid broader concerns about how quickly militaries can adopt AI and what guardrails they should apply. Helen Toner, the interim executive director of Georgetown University’s Center for Security and Emerging Technology, said the Pentagon’s latest contracts come during a period of anxiety over potential over-reliance on the technology in operational settings.
Toner said a lot of modern warfare already involves commanders making complicated choices from monitoring screens, and that AI can help by summarizing information or scanning surveillance feeds to identify potential targets. At the same time, she said questions remain about the appropriate level of human involvement, along with the risk of errors and how operators would be trained to use the tools effectively.
She warned that the Pentagon needs to balance rapid deployment for effectiveness with training and with steps to prevent over-trust. Toner said the challenge is to roll out these tools quickly enough to provide strategic advantage while also training operators and ensuring they do not assume the systems are always right.
The report described how those worries resemble sticking points raised by Anthropic in its dispute with the administration, including the company’s stance that it sought assurances in its contract that the military would not use its technology in fully autonomous weapons and would not use it to surveil Americans. It also said Defense Secretary Pete Hegseth responded that Anthropic would have to allow any uses the Pentagon deemed lawful.
The companies contracting with the Pentagon were also described in the context of the broader pace of adoption. The Pentagon said personnel were already using AI capabilities through its official platform, GenAI.mil, and that warfighters, civilians and contractors are putting the capabilities to practical use “right now,” cutting certain tasks from months to days.
The Pentagon said the expanding AI capabilities will “give warfighters the tools they need to act with confidence and safeguard the nation against any threat.” Toner said in many cases AI is used much like it is in civilian life: handling routine tasks that might otherwise take humans hours or days, such as forecasting maintenance needs for equipment or helping plan how to move troops and gear efficiently.
Toner said AI also can support decisions about whether a drone feed shows civilian or military vehicles. She added, however, that people should not become overly dependent on the systems, citing a “phenomenon called automation bias,” in which people may assume machines perform better than they do.
Separately, Pentagon technology leadership addressed the rationale for broadening vendor participation. Emil Michael, the Pentagon’s chief technology officer, told CNBC that it would have been irresponsible to rely on only one company and said the Pentagon moved to include multiple providers after learning that one partner did not want to work in the way the Pentagon intended.
The report said that some of the named firms already worked with the military in classified environments, including Amazon and Microsoft, but that others such as Nvidia and Reflection are new to such work. It also noted that both Nvidia and Reflection make open-source AI models, which Michael described as a priority in providing an “American alternative” to China’s rapid development of AI systems.
OpenAI, which was named among the seven companies, confirmed in a statement Friday that it was the same agreement the company announced in early March. The company said it believes “the people defending the United States should have the best tools in the world.” The report also said one company’s agreement included language calling for human oversight in missions in which the AI systems act autonomously or semiautonomously, and said it required the tools to be used in ways consistent with constitutional rights and civil liberties.