Overview
- Anthropic rejected revised contract language and reaffirmed two red lines for Claude: no mass surveillance of Americans and no use in lethal weapons without a human in the loop.
- Pentagon officials set a 5:01 p.m. ET Friday deadline and warned of canceling contracts, designating Anthropic a supply‑chain risk, or potentially invoking the Defense Production Act.
- Undersecretary Emil Michael publicly attacked CEO Dario Amodei, while spokesman Sean Parnell insisted the military seeks only "all lawful uses" and denies plans for mass surveillance or fully autonomous weapons.
- Claude is one of only two generative models on classified networks, and defense and industry sources say offboarding or replacing it could take months or longer due to deep AWS integrations.
- Rival labs hold Pentagon deals and some accept an "all lawful uses" standard, yet workers at Google and OpenAI and former DoD AI chief Jack Shanahan voiced support for Anthropic, and Michael said he remains open to talks before the deadline.