Claude under fire: Pentagon pressures Anthropic over military use

Anthropic

• Pentagon urges Anthropic to relax Claude AI limits for defense use
• Anthropic faces threat of being labeled a supply chain risk
• Claude AI's surveillance and weapons restrictions cause dispute

Did our AI summary help?

The United States Department of Defense has summoned Dario Amodei, CEO of AI developer Anthropic, for a critical meeting at the Pentagon amid a growing dispute over the use of its Claude model for national defence applications.

According to reporting from Axios and Reuters, Defence Secretary Pete Hegseth is pushing Amodei to soften or remove conditions that currently limit how the military can deploy Claude — particularly around mass surveillance of Americans and the development of fully autonomous weapons — or face serious consequences, including possibly being cut out of Pentagon workflows entirely.

A high-stakes ultimatum

Anthropic signed a $200 million contract with the DoD last year to provide its AI tools, including a customised “Claude Gov” model, for classified operations. Claude is said to be the only frontier AI system currently approved for use on defence and intelligence networks.

But the relationship has frayed. Pentagon officials want fewer restrictions on how Claude can be used — arguing that defence personnel should have broader flexibility within classified systems. Anthropic, by contrast, has sought to maintain guardrails that prevent its models from being used for intrusive domestic surveillance or in weapons that operate without human involvement.

A senior defence official told Axios that the upcoming meeting is not intended as a cordial discussion, but rather as an ultimatum: Acquiesce to the Pentagon’s requirements or be forced out of defence work.

Supply chain risk label looms

One of the Pentagon’s most severe threats is to label Anthropic a “supply chain risk” — a designation usually reserved for foreign adversaries and entities perceived as threats to national security. If applied, the tag could void Anthropic’s existing contract with the DoD and require other defence partners and contractors to drop Claude from their systems.

Such a designation would be unprecedented for a U.S. AI company and could have ripple effects across government and private sector deployments that rely on Claude technology.

Despite the tension, Anthropic maintains that talks remain constructive and in good faith, according to a company spokesperson quoted by Reuters. The firm emphasises its commitment to supporting national security while upholding safety principles.