Pentagon expands AI deals with major firms

What's happened

The Defense Department has broadened AI access across multiple contractors, allowing use of promised technologies for any lawful government purpose. The move comes amid a clash with Anthropic and amid ongoing litigation, while other big tech players like Google, OpenAI, xAI, and AWS are increasingly integrated into national security work.

What's behind the headline?

Analysis

The Pentagon is moving toward a diversified AI arsenal, reducing dependency on any single vendor and pushing a common standard across contractors.
Anthropic remains resistant to relaxing guardrails, while its Mythos model is increasingly viewed as a national-security asset despite ongoing disputes with the administration.
The deals raise questions about oversight, domestic surveillance, and the boundaries of autonomous capability; the government is balancing speed with risk controls.
Expect continued negotiations on contractual language, human-in-the-loop requirements, and potential limits on drone piloting and surveillance use.

tone:

This is a strategic shift toward an AI-enabled military posture, with clear implications for how technology firms interact with the U.S. government.

How we got here

The Pentagon has been pursuing broader use of AI across its forces since early this year, aiming to reduce vendor lock-in and accelerate decision‑making. Anthropic has resisted broad deployment beyond guardrails, while White House discussions with the company continue in the background. Google, OpenAI, xAI and others have struck agreements to provide models for classified networks, prompting negotiations over safeguards and human oversight.

Our analysis

New York Times — Julian E. Barnes has reported that the Defense Department has executed new agreements with multiple firms to enable their AI tools for any lawful government purpose, while Anthropic faces a supply chain risk designation and ongoing litigation. The report highlights friction with Anthropic over guardrails and the Pentagon's push for a unified standard. The New York Times coverage also notes that Google, OpenAI, xAI, AWS, Microsoft, Nvidia, and Reflection AI are part of the broader set of agreements. See also Sheera Frenkel's reporting on Google's deal and ongoing DOE/industry negotiations. The Japan Times summarizes the agreement's scope with respect to classified networks and government purposes.

Go deeper

How will the Pentagon enforce a single standard across all contractors?
What safeguards will accompany broad use of AI in classified environments?
How might Anthropic's court case influence future government deals?