Overview
- Defense Secretary Pete Hegseth labeled Anthropic a supply‑chain risk on March 3 and ordered a six‑month phase‑out across the Pentagon and its contractors, Reuters reports.
- Military staff, contractors and officials describe Claude as superior for current workflows and say some units are dragging their feet or planning to revert if the dispute is resolved.
- Claude was the first AI model cleared for classified networks and has supported operations related to Iran, underscoring how deeply it is embedded in targeting, planning and analysis tasks.
- Recertifying replacements for classified or military networks could take months, with some systems taking 12 to 18 months, leading to productivity losses and costly rebuilds, sources say.
- Palantir’s Maven Smart Systems includes prompts and workflows built with Claude Code that now must be reworked, and agencies are weighing pivots to OpenAI, Google or xAI under DoD directives to assess reliance.