Glossary

Guardrails.

the limits and safety measures placed around a model to keep its outputs within bounds.

the limits and safety measures placed around a model to keep its outputs within bounds. Related to alignment, the broader problem of making AI do what we actually intend.

Updated 2026-06-03