Claude's Constitution
On 2026-01-22, Anthropic released a new constitution for Claude -- an approximately 80-page document released under a Creative Commons CC0 1.0 deed (public domain, freely usable by anyone), marking a fundamental departure from Anthropic's original 2023 constitutional-AI approach.
The four values, in priority order
Anthropic wants all current Claude models to be, in this explicit order:
- Broadly safe -- not undermining appropriate human mechanisms to oversee AI during the current phase of development.
- Broadly ethical -- being honest, acting according to good values, and avoiding actions that are inappropriate, dangerous, or harmful.
- Compliant with Anthropic's guidelines.
- Genuinely helpful.
Explanation over instruction
The new constitution prioritizes explaining underlying principles over listing specific rules -- the framework is designed so that Claude could construct any rule Anthropic might come up with by understanding the reasoning behind it, rather than memorizing a static rulebook. It is described as "a detailed description of Anthropic's vision for Claude's values and behavior" and functions as a real input into the model training process, not merely a public communications document.
"Claude's constitution is a living document and a continuous work in progress. This is new territory, and Anthropic expects to make mistakes and hopefully correct them along the way."
Sources: anthropic.com/constitution, anthropic.com/news/claude-new-constitution.