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

# PII/PHI Detection Guardrail

> Detect and redact PII or PHI in LLM inputs, outputs, and MCP tool calls using the built-in TrueFoundry guardrail.

This guide explains how to use TrueFoundry's built-in **PII/PHI Detection** guardrail to detect and redact personally identifiable information or protected health information in LLM interactions and MCP tool invocations.

<Note>
  **Implementation:** This guardrail is powered by **Azure AI Language PII Detection** and runs on **TrueFoundry-managed** infrastructure — no third-party API keys or setup required. For vendor-hosted PII services using your own credentials (Azure PII, Google Model Armor, and others) or self-hosted redaction (for example via [Custom Guardrails](/docs/ai-gateway/custom-guardrails)), see [Supported Guardrails](/docs/ai-gateway/guardrails-overview#supported-guardrails) and [Guardrails Overview](/docs/ai-gateway/guardrails-overview).

  PII/PHI Detection can be applied to **all four guardrail hooks**: LLM Input, LLM Output, MCP Pre Tool, and MCP Post Tool—providing comprehensive PII/PHI protection across your entire AI workflow.
</Note>

## What is PII Detection?

PII Detection is a built-in TrueFoundry guardrail that identifies and redacts personally identifiable information or protected health information from text content. It is powered by **Azure AI Language PII Detection** under the hood and understands context to detect PII/PHI with high accuracy even when the data doesn't follow a strict format. The guardrail is fully managed by TrueFoundry — no external credentials or setup required.

### Key Features

1. **Context-Aware Detection**: Detects PII/PHI with contextual understanding, providing higher accuracy and fewer false positives than simple pattern matching.

2. **Configurable Entity Categories**: Select which types of PII/PHI to detect from a comprehensive set of entity categories from basic identifiers like names and emails to financial and identity document numbers.

3. **Automatic Redaction**: Operates in mutate mode, detected PII/PHI is automatically redacted in the content, allowing the request to continue with sensitive data removed.

## Adding PII/PHI Detection Guardrail

<Steps>
  <Step title="Navigate to Guardrails">
    Go to the AI Gateway dashboard and navigate to the **Guardrails** section.
  </Step>

  <Step title="Create or Select a Guardrails Group">
    Create a new guardrails group or select an existing one where you want to add the PII/PHI Detection guardrail.
  </Step>

  <Step title="Add PII Detection Integration">
    Click on **Add Guardrail** and select **PII/PHI Detection** from the TrueFoundry Guardrails section.

    <Frame caption="Select PII Detection from TrueFoundry Guardrails">
      <img src="https://mintcdn.com/truefoundry/yRoKH_fkKi2nPtuV/images/guardrail-1.jpeg?fit=max&auto=format&n=yRoKH_fkKi2nPtuV&q=85&s=9ff04ad219001f1bfc31959b9ac261da" alt="TrueFoundry guardrail selection interface showing PII/PHI Detection option" width="1280" height="793" data-path="images/guardrail-1.jpeg" />
    </Frame>
  </Step>

  <Step title="Configure the Guardrail">
    Fill in the configuration form:

    * **Name**: Enter a unique name for this guardrail configuration (e.g., `pii-detection`)
    * **PII Categories**: Select the entity categories you want to detect or select all categories(Default).
    * **Enforcing Strategy**: Choose how violations are handled

    <Frame caption="Configure PII/PHI Detection with entity categories">
      <img src="https://mintcdn.com/truefoundry/SUZtx2F6kgBTQYk8/images/guardrail-tfy-pii-config.png?fit=max&auto=format&n=SUZtx2F6kgBTQYk8&q=85&s=a5dea066bbcbd4d2780b2e093c741bb1" alt="PII/PHI Detection configuration form showing entity category selection" width="1516" height="1172" data-path="images/guardrail-tfy-pii-config.png" />
    </Frame>
  </Step>

  <Step title="Save the Configuration">
    Click **Save** to add the guardrail to your group.
  </Step>
</Steps>

## Configuration Options

| Parameter              | Description                                              | Default   |
| ---------------------- | -------------------------------------------------------- | --------- |
| **Name**               | Unique identifier for this guardrail                     | Required  |
| **Operation**          | `mutate` only (detects and redacts PII)                  | `mutate`  |
| **Priority**           | Execution order for mutate guardrails (lower runs first) | `1`       |
| **Enforcing Strategy** | `enforce`, `enforce_but_ignore_on_error`                 | `enforce` |
| **PII Categories**     | Array of entity categories to detect                     | Required  |

<Note>
  PII Detection only supports **mutate** mode — it always redacts detected entities. See [Guardrails Overview](/docs/ai-gateway/guardrails-overview#operation-modes) for details on Operation Modes and Enforcing Strategy.
</Note>

## How It Works

When content passes through the guardrail, detected PII entities are replaced with asterisks in the output, preserving the rest of the text.

**Example**:

```
Input: "Call our office at 312-555-1234, or send an email to support@contoso.com"
Output: "Call our office at ************, or send an email to *******************"
```

```
Input: "My name is John Smith and I live at 742 Evergreen Terrace, Springfield"
Output: "My name is ********** and I live at *************************************"
```

## Use Cases

### Recommended Hook Usage

| Hook              | Use Case                                                       |
| ----------------- | -------------------------------------------------------------- |
| **LLM Input**     | Redact PII from user messages before they reach the LLM        |
| **LLM Output**    | Redact PII from LLM-generated responses                        |
| **MCP Pre Tool**  | Remove PII from tool parameters before invocation              |
| **MCP Post Tool** | Redact PII from tool outputs (database results, file contents) |