Home › Research Papers › Constitutional AI (2022)
Constitutional AI (2022)
Research Paper
Alignment
Anthropic's method: models critique themselves against written principles.
You're looking at a research-paper record: a landmark paper and what it introduced to the field.
Instead of scaling human labeling of harms, Constitutional AI has the model critique and revise its own outputs against an explicit constitution, then trains on the result (RLAIF). It powers Claude's character, made the principles inspectable, and pointed toward oversight that scales beyond human review.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Introduced in
Basis for
Sources
Record reviewed as of 2026-01-15 — see Methodology.