Overview
- Defense Secretary Pete Hegseth is described as close to severing business with Anthropic and considering a rare supply‑chain risk designation, according to Axios citing a senior Pentagon official.
- The proposed designation would force Pentagon contractors to certify they do not use Anthropic’s Claude, a step typically reserved for foreign adversaries.
- Claude is the only commercial AI model on U.S. classified military networks and was used in the January Maduro operation, making any abrupt replacement difficult, sources told Axios.
- At issue is the Pentagon’s push for access to leading models for “all lawful purposes,” while Anthropic maintains hard limits on mass domestic surveillance and fully autonomous weapons.
- OpenAI, Google, and xAI have agreed to relax safeguards for unclassified military systems, and a contract worth up to about $200 million with Anthropic is at risk as negotiations continue.