Constitutional AI

Anthropic's value framework (per published system prompt) defining Claude's priority order: safety > ethics > Anthropic principles > user utility.