Overview
- Defense Secretary Pete Hegseth demanded unrestricted military use of Claude by 5 p.m. Friday, warning of contract termination, a supply‑chain risk designation, or invocation of the Defense Production Act.
- Officials sent a best‑and‑final offer to Anthropic on Wednesday night, with no agreement reported and negotiations still active.
- Anthropic reaffirmed two red lines—no mass surveillance of Americans and no fully autonomous lethal targeting—even after earlier agreeing to missile and cyber defense uses.
- Claude remains embedded on classified systems via Palantir and has been described by defense officials as uniquely capable, while xAI’s Grok was cleared this week for classified use for any lawful mission.
- On the same day tensions escalated, Anthropic softened parts of its Responsible Scaling Policy by dropping prior hard‑stop pledges, a shift critics say weakens safety guardrails.