Safety · advanced
What is Constitutional AI?
A plain-English explanation of Constitutional AI (Constitutional AI) — what it means, why it matters, and how it is used in AI.
Constitutional AI
Constitutional AI
Constitutional AI is a training technique developed by Anthropic in which an AI model is trained to follow a set of written principles by critiquing and revising its own outputs against those principles.
"Rather than having humans label millions of harmful responses, Constitutional AI has the model review its own outputs against principles like "do not assist with creating weapons.""
Also known as: Constitutional AI, CAI, self-critique training
Why does Constitutional AI matter?
Constitutional AI is Anthropic's core safety training technique and is fundamental to how Claude is trained.
Practice this term
The best way to remember Constitutional AI is to practice unscrambling it. AI Terminology Scrambler uses spaced repetition to help you learn and retain AI vocabulary in just a few minutes a day.
Practice Constitutional AI now →
Related AI terms