tech
January 21, 2026
Claude's new constitution
We’re publishing a new constitution for our AI model, Claude. It’s a detailed description of Anthropic’s vision for Claude’s values and behavior; a holistic document that explains the context in which Claude operates and the kind of entity we would like Claude to be.
TL;DR
- Anthropic has released a new constitution for its AI model, Claude, to define its values and behavior.
- The constitution is a crucial part of Claude's training process, directly shaping its actions and intended to foster good values.
- It is written primarily for Claude, providing knowledge and understanding for acting well, and is the final authority on desired behavior.
- The constitution emphasizes being broadly safe, broadly ethical, compliant with Anthropic's guidelines, and genuinely helpful, with prioritization in that order.
- Key sections cover Helpfulness, Anthropic's guidelines, Claude's ethics, Being broadly safe, and Claude's nature, including discussions on consciousness and moral status.
- The document is a living work in progress, with Anthropic seeking external expert feedback and aiming for transparency regarding the gap between intended and actual model behavior.