Pentagon signs new AI contracts with tech giants for classified networks

The Pentagon has moved commercial artificial intelligence deeper into the military’s most sensitive systems, approving new contracts that will let major tech firms work inside classified network environments at Impact Level 6 and Impact Level 7. The agreements mark a sharper turn from testing tools on the edges of defense work to embedding AI inside the information channels that support targeting, logistics, surveillance, and command decisions.

The Defense Department said the new deals accelerate its transformation toward an “AI-first fighting force,” language that reflects a broader doctrinal shift already taking shape in Washington. A January 2026 Pentagon AI strategy called for the department to become an “AI-first” warfighting force, and a separate modernization document said the department was executing an AI strategy to make it an “AI-first warfighting force.” Defense officials have also said the department has been investing in AI for more than 60 years.

Reporting identified Amazon Web Services, Google, Microsoft, OpenAI, SpaceX, NVIDIA, and Reflection among the companies involved, although accounts differed on whether the Pentagon had named seven firms or signed eight contracts. The department did not say when the systems would be available on classified networks or how much the companies would be paid, leaving the scale and timing of the rollout unclear.

The move builds on the secure but unclassified GenAI.mil platform, which was stood up in December 2025, and on an earlier surge of spending. In July 2025, the Pentagon awarded up to $800 million in AI contracts to Google, OpenAI, Anthropic, and xAI, with each company eligible for up to $200 million in work. The latest agreements suggest the department is moving beyond pilot programs toward wider operational use across military systems.

That shift has intensified the accountability debate. Anthropic reportedly objected to unrestricted access and raised concerns that its models could be used for fully autonomous weapons or surveillance of Americans. Pete Hegseth and the Trump administration then moved to bar or blacklist Anthropic from some government work, and Anthropic sued. OpenAI later said the Pentagon agreement announced Friday was the same deal it had announced in early March 2026, effectively replacing Anthropic in at least some classified environments.

Supporters argue the technology could speed battlefield target identification, improve strike decisions, sharpen logistics and weapons maintenance, and help commanders synthesize huge volumes of data into usable situational awareness. Critics and civil-liberties advocates warn that the same systems could widen surveillance, erode privacy, and push military decisions closer to automation. One report said the Pentagon’s agreements included human oversight language for autonomous or semi-autonomous missions, a sign that officials are trying to preserve a human hand on the trigger even as commercial AI becomes more central to national defense.