Pentagon bans Anthropic tech after dispute over Claude safeguards

In short: Anthropic and the Pentagon are still publicly clashing over rules for how the Claude AI model can be used, after the US government banned Anthropic’s technology across federal agencies.

What happened

Anthropic says it will only provide its Claude model with two firm limits in place. The company says Claude should not be used for mass surveillance of Americans, and it should not be used in fully autonomous weapons without a human in control.

The Pentagon, led by Defense Secretary Pete Hegseth, has pushed back. Officials say the Department of Defense needs access for “any lawful use,” and they argue that mass surveillance would be illegal anyway. Anthropic says the Pentagon’s proposed contract language leaves loopholes that could let the limits be bypassed.

The disagreement escalated in early 2026. Hegseth issued an AI contracting memo in January. In early March, he met Anthropic CEO Dario Amodei, set a deadline for full access, and sent what the Pentagon called a “best and final offer.”

On February 27, 2026, Hegseth designated Anthropic as a national security supply chain risk, which triggered a six month transition period away from its tools. President Trump also posted on Truth Social attacking Anthropic and announced a federal ban, even as Claude was reportedly still used in some military activity soon after.

Anthropic says it will challenge the designation in court and will help the government transition to alternatives. The company also says its limits have not disrupted work like intelligence analysis and cyber operations.

Why it matters

This fight is about who sets the rules for powerful AI tools used by the government. Think of it like a car with speed limiters. The Pentagon wants to decide when to remove them, and the maker wants them to stay on for safety and civil rights reasons.

Source: TechCrunch AI