Constitutional AI

A training methodology where AI models are guided by a set of principles or a constitution, enabling the model to self-critique and revise outputs to align with specified rules.

Training an AI with a written set of rules (a "constitution") it must follow. The AI learns to check its own responses against these rules and correct itself; like having an internal code of conduct.

Constitutional AI represents an approach to embedding governance principles directly into AI model training. Understanding this technique helps governance teams assess whether AI providers are taking alignment and safety seriously.