Redact / Masking / Placeholder

Redaction is the strictest Guardian Layer output style. Instead of replacing sensitive values with readable alternatives, it hides detected values with masking marks.

Use this page when you want to understand the difference between readable transformation and visible masking.

Redact vs Masking vs Placeholder

These terms are closely related, but they are useful in slightly different ways.

Term	What it means in this context	Example output
Redact	Hide the detected sensitive value.	`Dr. *** called ***.`
Masking	The visible masking style used by redaction.	`*****`
Placeholder	A label-style replacement that describes the entity type.	`[PERSON]`, `[PHONE_NUMBER]`

In the current Guardian Layer API, redaction is selected with masking_type: "redact" on the text de-identification endpoint.

Placeholder-style output is useful for documentation and product design discussions, but it should not be treated as a separate request mode unless it is exposed by the API.

When to Use Redact

Use redact when you want to:

Clearly show that sensitive data was removed or hidden.
Avoid generating replacement values that could be mistaken for real data.
Review sensitive content in a compliance or audit workflow.
Build a safer first pass before deciding whether transformed output is appropriate.

Use transform instead when you need readable de-identified text that still feels natural.

Endpoint

POST /api/v1/deidentify/text/proprietary/outputs

Redaction uses the same endpoint as Transform. The difference is the masking_type value.

Request

Set masking_type to redact.

{
  "text": "Dr. Emily Chen called Alex at 555-0100 about the appointment in Auckland.",
  "domain": "General",
  "pca_dimension": 2,
  "masking_type": "redact",
  "pii_entities": ["PERSON", "PHONE_NUMBER", "LOCATION"]
}

Parameters

Parameter	Type	Required	Description
`text`	string	Yes	The input text you want to de-identify.
`domain`	string	No	Detection domain. Supported values are `General`, `Medical`, `Finance`, and `Custom`. Defaults to `General`.
`pca_dimension`	integer	No	Detection setting used by the endpoint. Supported values are `1`, `2`, and `3`. Defaults to `2`.
`masking_type`	string	Yes	Use `redact` to replace detected sensitive values with masking marks.
`pii_entities`	array	No	Limits which entity categories are redacted. Omit this field, pass an empty list, or pass `["ALL"]` to process all supported entities.

For supported entity values, see Supported Entities.

Response

The response shape is the same as Transform. In redaction mode, the returned outputs may contain the same redacted text.

{
  "original_text": "Dr. Emily Chen called Alex at 555-0100 about the appointment in Auckland.",
  "outputs": [
    {
      "id": 1,
      "text": "Dr. ***** called ***** at ***** about the appointment in *****."
    },
    {
      "id": 2,
      "text": "Dr. ***** called ***** at ***** about the appointment in *****."
    }
  ],
  "mode": "PROPRIETARY_OUTPUTS",
  "meta": {
    "masking_type": "redact",
    "domain": "General",
    "pca_dimension": 2
  }
}

Choosing the Right Output Style

Need	Recommended mode
Make sensitive content unreadable	`redact`
Preserve a natural-looking sentence	`transform`
Create sample data for demos or AI workflows	`transform`
Make removed values obvious to reviewers	`redact`
Show entity labels such as `[PERSON]`	Placeholder-style output, if supported by your implementation

Notes

Redaction hides detected sensitive values with masking marks.
Entity filtering works the same way as Transform.
The API response still uses the outputs array, even when both redacted outputs are identical.
Redaction is usually easier to audit because hidden values are visually obvious.
Placeholder-style labels should be documented separately if they become an exposed API option.

Redact vs Masking vs Placeholder​

When to Use Redact​

Endpoint​

Request​

Parameters​

Response​

Choosing the Right Output Style​

Notes​

Next Steps​