Pentagon moves to restrict Anthropic after dispute over Claude use

In short: The US Department of Defense is pressuring Anthropic to loosen limits on how the military can use its Claude AI, and has threatened to cut ties when Anthropic refused.

What happened

The US Department of Defense (DOD) and Anthropic, the company behind Claude, are in a public dispute over how the military is allowed to use the company’s AI models.

According to Wired, Defense Secretary Pete Hegseth said on February 27, 2026 that the DOD would label Anthropic a “Supply-Chain Risk to National Security.” In simple terms, that is like putting a supplier on a “do not buy” list for defense contractors, after a six-month transition period.

The clash centers on contract terms. The DOD wanted “unrestricted” use for “any lawful purpose.” Anthropic CEO Dario Amodei said some uses are off limits, including fully autonomous weapons (systems that can pick and strike targets on their own) and domestic mass surveillance (using AI to watch large numbers of people inside the US).

Wired also reports the DOD accused Anthropic of being able to manipulate its models during a war. Anthropic executives said that is not possible. Separately, the report says OpenAI signed a DOD contract soon after, including safeguards that claim to ban intentional domestic surveillance of US persons, though critics questioned how those rules would be defined and enforced.

Why it matters

This fight is about who sets the rules for powerful AI tools when national security is involved. The outcome could shape whether AI is used more like a decision aid for humans, or more like an automated system that can act at scale, including in surveillance.

Source: Wired