The Pentagon has reached new agreements with four technology companies to broaden the use of advanced artificial intelligence tools on classified military networks, the Department of Defense said — a report Bloomberg published on Friday confirmed. The new partners named in the announcement are Nvidia Corp., Microsoft Corp., Reflection AI Inc., and Amazon.com Inc., and the agreements allow AI capabilities to be used for "lawful operational use."
Officials involved in internal deliberations spoke to the press anonymously. According to the Defense Department, the new arrangements give the Pentagon substantial latitude to deploy high-performance AI technologies on classified and top-secret networks, including for tasks such as target designation support.
Changes after the Anthropic dispute
The latest agreements relax some of the restrictions that Anthropic PBC had sought to impose — restrictions that contributed to the collapse of a Pentagon deal with Anthropic earlier this year. Anthropic opposed use of its technology for mass domestic surveillance and fully autonomous weapons; the Pentagon declined those proposed limits and subsequently accelerated efforts to bring on other AI vendors willing to accept broader terms.
Defense officials say the aim is to avoid dependence on a single supplier or a single set of use restrictions. Nvidia's agreement is notable for the greater freedom it grants: the company agreed not to impose usage or licensing terms more restrictive than U.S. law and to provide “full and effective access” to its capabilities, including for the development of autonomous weapon systems. The Pentagon said use would still be consistent with U.S. citizens' constitutional rights, but it did not detail specific oversight mechanisms in the announcement.
Broader context and pace of adoption
Since autumn 2025 the Pentagon has tried to expand permitted AI uses; the Defense Department said a number of earlier deals — including arrangements with OpenAI, Google, and SpaceX — reflect the same objective. The announcement references seven companies in total and signals an effort to accelerate the U.S. military's transition to AI-enabled combat capabilities.
Bloomberg also reported that Anthropic's Claude system is currently used in operations involving Iran, according to defense sources; the Pentagon has set a six-month timeline to replace the Claude system and is exploring alternatives.
Reactions from companies and the defense side
During negotiations, Tim Barrett, a spokesperson for Amazon Web Services, said AWS has supported the U.S. military for more than a decade and intends to remain involved in Defense Department modernization efforts by developing AI solutions that assist critical missions. Nvidia did not immediately comment on the new agreement; a Microsoft spokesperson declined to comment, and a Reflection representative was not available, Bloomberg reported.
The dispute has also played out politically. Secretary of Defense Pete Hegseth told Congress that AI does not make lethal decisions and emphasized that humans make the final calls; he also criticized Anthropic's leadership, according to Bloomberg.
Risks and civil society concerns
Civil organizations have warned about the risks of military AI use. Key concerns include system errors, the danger of automation bias where humans over-rely on machine recommendations, and ethical questions around developing fully autonomous weapons. The struggle between the Pentagon and Silicon Valley over limits on military AI underscores ongoing tensions about risk, accountability, and acceptable use.
International and political implications
The technological competition favors actors with comparable industrial and research capacities. Bloomberg noted this dynamic in references to recent exchanges between Washington and Kyiv: President Donald Trump reportedly told Ukrainian President Volodimir Zelenszkij that the U.S. does not need Ukraine's drone capabilities for operations in the Middle East, asserting "We know more," a remark pointing to U.S. technological advantages.
Overall, the Pentagon's new agreements with major tech firms mark a significant step in building AI-based military capabilities, while intensifying debate over their legal, ethical, and operational boundaries.
