> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cognigy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Guardrails

> Guardrails provide safety and data privacy instructions for agents.

> [!NOTE]
>
> This feature is in beta. We encourage you to try it out and provide us with feedback.

*Guardrails* are predefined safety and data privacy settings you can activate for an agent. Use guardrails to guide your agent to comply with content and data privacy policies.

## Key benefits

* **Swift configuration**. You can quickly apply content safety policies when creating agents.
* **Personal data handling**. You can customize how your agent handles contact profile information to comply with data privacy regulations.

## Restrictions

* You can't create custom guardrails. To enforce additional rules, add them to the agent's instructions, or adjust the description of a specific [tool](https://docs.cognigy.com/ai/agents/develop/tools.md) or [skill](https://docs.cognigy.com/ai/agents/develop/skills.md).

## How to use guardrails

You can configure guardrails when you [create an agent](https://docs.cognigy.com/ai/agents/overview#create-an-agent) or edit an existing one:

1. Go to **Design > Agent**.

2. Select the agent.

3. Go to the **Guardrails** tab and configure the following.

   * In the **Safety settings** section, activate the toggles you need under **Safety instructions**.

     * **Avoid harmful content** — prevents the agent from generating content that could be harmful, offensive, or abusive to end users.
     * **Avoid ungrounded content** — prevents the agent from generating content that is based on speculation or unsupported claims, ensuring it is reliable and verifiable.
     * **Avoid copyright infringements** — prevents the agent from generating content that violates intellectual property rights or uses copyrighted material without authorization.
     * **Prevent jailbreak and manipulation** — prevents attempts to bypass security measures or manipulate the agent into producing unauthorized or unsafe content.

     Activating a safety setting adds it to the agent's prompt, which can increase token usage.

     All safety settings are activated by default.

     Guardrails guide the model but don't guarantee its behavior. Occasional unintended outputs can still occur.

   * In the **Data privacy** section, select an option in **Contact profile information** to define how your agent uses contact profile information to personalize responses.

     * **No profile fields** — the agent doesn't use any contact profile data. This option is selected by default.
     * **Selected profile fields** — displays the **Profile fields** list, where you select the fields the agent can access.
     * **All profile fields** — the agent uses all fields from the contact profile.
     * **Profile memories only** — the agent uses only the [memories](https://docs.cognigy.com/ai/agents/analyze/contact-profiles.md#default-profile-schema) field from the contact profile.

4. Click **Save changes**.

## More information

* [Manage Agents](https://docs.cognigy.com/ai/agents/develop/manage-ai-agents.md)
* [Contact profiles](https://docs.cognigy.com/ai/agents/analyze/contact-profiles.md)
* [Persona](https://docs.cognigy.com/ai/agents/develop/persona.md)