Anthropic got until Friday to save its $200M Pentagon contract or be treated like a foreign adversary.
In the last few days, Anthropic did everything to show good citizenship:
- Exposing Chinese labs running industrial-scale model distillation
- Releasing Claude Code Security
- Publishing a new Responsible Scaling Policy v3.0
None of this is coincidental.
But the Pentagon wants unrestricted model access to fight wars. Claude's current terms don't allow domestic surveillance or autonomous lethal operations.
Friday we'll find out if principles survive a $200M ultimatum.