The U.S. Department of Defense has completed agreements with eight major technology companies to deploy AI capabilities in classified settings, escalating the Pentagon's reliance on commercial AI in national security operations. The eight: Alphabet (GOOGL), Microsoft (MSFT), Amazon (AMZN), Oracle (ORCL), Nvidia (NVDA), OpenAI, SpaceX's xAI, and startup Reflection AI.
Notably absent from the list: Anthropic, the only major AI lab to reject Pentagon contracts outright. Anthropic CEO Dario Amodei insisted on guardrails against domestic mass surveillance and autonomous weapons deployment. Defense Secretary Pete Hegseth responded by calling Amodei an ideological opponent in a public statement—a rare personal rebuke that highlights the philosophical fracture dividing Silicon Valley.
The Pentagon's pivot reflects a shift in thinking about AI security. Rather than demanding closed-source, government-built systems, the Defense Department is now comfortable deploying open-source models from vendors like Reflection AI and open-source frameworks supported by Nvidia. The logic: open-source models are harder for foreign adversaries to penetrate, and they offer faster iteration cycles.
"We can move faster with commercial partners than internal R&D."
The eight-vendor agreement also signals confidence in the mega-cap cloud providers (Google, Microsoft, Amazon, Oracle). All have demonstrated ability to run classified workloads on secure infrastructure. Nvidia's inclusion underscores the Pentagon's view that semiconductor superiority is the real gating factor in AI competition with China.
For Anthropic, the rejection of Pentagon money may prove costly. The company is fighting legal cases to overturn the Defense Department ban, but the political winds are blowing against regulatory guardrails in the AI space. Anthropic's bet is that long-term reputation and ethical positioning will pay off. The market is currently skeptical.