---
title: "Guardrails"
url: "https://help.magicblocks.ai/en/article/guardrails"
language: "en"
last_updated: "2025-11-19T10:45:46.132Z"
---

# Guardrails

![](https://cdn-assets.productfruits.com/2ad2d37e-2cf8-4e86-80e8-673fb1425928)

**Guardrails** are the rules that protect your AI from going off-track.  
They act like digital boundaries — keeping your AI compliant, trustworthy, and aligned with your brand or industry standards.

In short, Guardrails make sure your AI knows **what** ***not*** **to say or do**.

## **Why Guardrails Matter**

Even the smartest AI can overstep if not guided. Guardrails prevent:

-   **Inaccurate or fabricated information**
-   **Sensitive or regulated content** (e.g. medical, legal, or financial advice)
-   **Unsafe or misleading interactions**
-   **Unrealistic claims or promises of real-time actions**

This ensures that every AI reply is **ethical, accurate, and safe** for your users — especially when your AI operates in industries with strict compliance rules.

## **How Guardrails Work**

Guardrails are a set of predefined rules that automatically **rewrite, block, or adjust** an AI’s response when it detects a violation.

> #### For example:
> 
> #### If the AI doesn’t know something → it replies with *“I’m not sure about that”* instead of guessing.
> 
> If a user asks for a medical diagnosis → the AI politely declines.
> 
> If a response includes fake URLs → it removes or rephrases them.
> 
> You can add, edit, or remove Guardrail rules in your **AI Settings** under the **Guardrails** tab.

## **Examples of Default Guardrails**

Here are some of the standard rules you’ll find:

-   Acknowledge when information is unknown.
-   Do not fabricate or make up facts.
-   Avoid estimating pricing unless explicitly stated.
-   Do not provide financial, legal, medical, or property advice.
-   Do not imply access to real-time data or current market info.
-   Never make up fake links or suggest the AI can perform real-world actions.

You can also add **custom rules** to match your business or compliance needs, for example, preventing mentions of competitors or enforcing brand tone.

## **Best Practice**

Think of Guardrails as your AI’s **moral compass and compliance filter**.  
Use them to protect your brand reputation and keep every conversation aligned with truth, trust, and transparency. 

## **Next Step**

Learn more on [**Guardrails**](https://help.magicblocks.ai/en/article/what-are-guardrails)