Overview
- Defense Secretary Pete Hegseth told CEO Dario Amodei that Anthropic must allow Claude for all lawful military uses by Friday or risk contract termination, a supply‑chain‑risk label, or action under the Defense Production Act.
- Claude is currently embedded on classified DoD networks under a roughly $200 million agreement, though officials say xAI’s Grok is now cleared for classified settings and other providers are nearing approval.
- Anthropic maintains two red lines for its AI: no fully autonomous weapons targeting and no mass domestic surveillance of Americans, while Pentagon officials insist tools must be available for any lawful purpose.
- Separately, Anthropic alleges DeepSeek, Moonshot AI and MiniMax used about 24,000 fraudulent accounts to generate over 16 million Claude exchanges for distillation, with MiniMax responsible for more than 13 million.
- The company says it traced the activity using IP correlations, metadata and proxy-reseller signals, shared its findings with U.S. authorities, and notes the accused firms have not publicly responded in these reports.