[Glossary](https://vatan.one/glossary)/**Guardrail**

# What is Guardrail?

A rule applied to a prompt or response before it is allowed through.

Typically redaction of identifiers, secret detection, prompt-injection screening or output classification. Applied at the gateway, not in the application, so it holds for every team, including those who did not read the policy.

Related

[Prompt injection](https://vatan.one/glossary/prompt-injection)[AI gateway](https://vatan.one/glossary/ai-gateway)

See how this works in practice

The console shows the mechanism rather than describing it.

[Products](https://vatan.one/products)

Source: https://vatan.one/glossary/guardrail
