The Pentagon said it has reached agreements with seven AI companies to allow their advanced tools to operate on Defense Department classified networks, while leaving Anthropic off the approved list amid concerns over guardrails and supply-chain risk.
SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft and Amazon Web Services will be integrated into the department’s secret and top-secret environments, enabling military access to their products for sensitive work. Reflection, a smaller AI company that raised $2 billion in October, is backed by 1789 Capital, a venture firm in which Donald Trump Jr. is a partner and investor.
Anthropic — previously widely used inside the Defense Department — was designated earlier this year as a “supply-chain risk” and barred from Pentagon networks and many contractors. The company and the department are involved in litigation following that March designation. Officials said Anthropic’s exclusion also reflects unresolved questions about appropriate guardrails for military use of its models.
Following the Anthropic dispute, the Pentagon has moved to speed up onboarding of newer AI providers into secret and top-secret data environments, shrinking approval timelines to under three months from the 18 months or longer it once took. Officials say broadening the vendor pool is intended to reduce the risk of overdependence on any single provider, often described as avoiding “vendor lock.”
Pentagon staff, former officials and IT contractors told Reuters they were reluctant to relinquish Anthropic’s tools, which some users consider superior, despite the department’s order to remove them within six months. AI adoption within the military has accelerated: the Pentagon’s main AI platform, GenAI.mil, has been used by more than 1.3 million Defense Department personnel after five months in operation.
Separately, sources said Google has struck a deal allowing the Defense Department to use its AI models for classified work. Defense Department Chief Technology Officer Emil Michael has reiterated that Anthropic remains a supply-chain risk. He also said Anthropic’s Mythos model — which U.S. officials and private companies flagged for its advanced cyber capabilities — represents a separate national security concern. Various companies and organizations have accessed a preview of Mythos to help shore up IT defenses, though it is unclear whether the Pentagon participated in that preview.
President Donald Trump recently said Anthropic was “shaping up,” a comment that could signal potential executive reconsideration of the company’s Pentagon status, though no reversal has been announced.