Glossary
Guardrails
Controls that constrain model inputs and outputs.
What it is
Guardrails validate and constrain what goes into and comes out of a model — input filtering, output schemas, safety classifiers, and refusal policies — to keep behavior safe and predictable.
Why it matters in production
Guardrails are the difference between a demo and a system you can put in front of real users: input validation, bounded agent loops, output schema checks, and treating all retrieved text as untrusted data.
In practice
Validating that a generated SQL query is read-only before running it, and rejecting inputs that attempt to override the system prompt.
Related terms
Production AI Notes
One practical AI engineering email each week
One concept, one architecture, one project idea, and one interview question — written for developers who want to build and ship real AI systems.