After a publicly messy fallout between the US Department of War (or the Pentagon) and the AI startup behind Claude, Anthropic, a new report from Bloomberg suggests that disagreement has now escalated into policy. The Trump administration has labelled Anthropic a “supply-chain risk,” putting its existing and future defence contracts in jeopardy.

Inside the Pentagon, the shift is already underway. “The Department is actively pursuing multiple LLMs into the appropriate government-owned environments,” said Cameron Stanley, the military’s chief digital and AI officer. “Engineering work has begun… and we expect to have them available for operational use very soon.”

The transition will not happen overnight. Anthropic’s tools, particularly its Claude model, had become embedded in certain workflows, including classified environments where few alternatives were previously approved. Replacing that kind of infrastructure takes time, especially when it connects to systems like Palantir’s Maven platform, which supports military operations.

Still, the Pentagon appears willing to absorb that friction. A six-month window has been set to phase out Anthropic’s technology and bring in other providers.

Those alternatives are already lining up. OpenAI and xAI have received approval to handle classified work, while Google is rolling out its Gemini AI tools across parts of the defence workforce, starting with unclassified systems. The contrast is hard to miss. Anthropic pushed for tighter guardrails. Others are stepping in to fill the gap, likely with different trade-offs around control, flexibility, and oversight.

This moment says as much about the industry as it does about one company. Governments are rushing to adopt advanced AI, but they are still negotiating the rules around its use in real time. What counts as a safeguard for one side can look like a limitation to another.

Anthropic is now challenging the decision in court, arguing that the government’s actions violate its rights and could damage its business. But inside the Pentagon, there is little indication of a reversal. “I don’t think there’s a scenario where this gets resolved in that way,” said Emil Michael, Under Secretary of Defense for Research and Engineering of the United States.

How Anthropic’s AI Talks With the Pentagon Unfolded: Full Timeline
Here’s the timeline of what happened.