· Updated
The constitution is written for Claude itself, not humans - it's the actual document used in the training pipeline. Claude also uses it to generate synthetic training data that reinforces the values it describes. The CC0 1.0 license means anyone can adapt the full text.
The priority hierarchy - safe > ethical > compliant > helpful - offers a useful mental model for thinking about AI decision-making tradeoffs.
We’re publishing a new constitution for Claude. The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process. https://t.co/CJsMIO0uej
