Feature

Data Redaction

Remove PII, PCI, and PHI from prompts, responses, or both at the gateway. Keep a record that the rule ran, without storing the raw values.

Redaction rule

Which entities were filtered and whether the prompt, response, or both were covered.

Scope

Prompt + response

Data

PII + PCI + PHI

Review

Events

New capabilities

What your team gains with Concentrate

Strip PII before it leaves

Remove selected sensitive entities (names, emails, card numbers, health identifiers) from the prompt before the request ever reaches a model provider.

Input

Card number

Redacted before provider request.

Output

Address

Filtered before the app receives the response.

Owner

Support key

Rule applied to this key's requests, whether a team key or personal key.

Filter what comes back

Redact sensitive fields from responses before they return to the app, so a model can't surface data into a UI or log that shouldn't have it.

Prove it ran

Keep a record that redaction fired on a request without storing the raw sensitive values, so your security team gets evidence instead of a promise.

Who Concentrate is designed for

How gateway-level redaction keeps sensitive data out of model traffic

Data redaction removes sensitive data like PII, PCI, and PHI before it reaches an AI provider, and before a response comes back to your app. Tools like AWS Bedrock Guardrails, Google Cloud DLP, and Microsoft Presidio each protect just one platform, or have to be built into every app by hand. Concentrate handles it as requests pass through, so one rule protects every model you use.

Prompt and response coverage

Rules can run on the prompt, the response, or both, so sensitive data is filtered on the way out to the provider and on the way back to the app.

Applied once, for every app

Redaction lives at the gateway, not in each codebase. New apps inherit the rules instead of each team writing and maintaining their own filtering.

Evidence without exposure

Redaction events record that filtering happened without storing the raw values, giving auditors proof while keeping the sensitive data out of logs.

Pairs with retention controls

Combine redaction with so sensitive fields are both removed and routed only to providers that don't retain data.

Feature basics

Frequently asked questions

Can Concentrate redact prompts and responses?

Yes. Redaction can run on the prompt, the response, or both, depending on how the rule is configured. That covers data leaving for a provider and data returning to your application.

Why use a gateway for sensitive data redaction?

A gateway applies redaction at the single point every model request passes through, instead of asking every app team to write and maintain their own redaction code. One rule covers current and future workloads, which is far easier to verify in a security review.

What kinds of sensitive data can be redacted?

Rules target sensitive entities such as PII (names, emails, phone numbers), PCI (card numbers), and PHI (health identifiers). You choose which entity types to redact and whether the rule applies to prompts, responses, or both.

One API for every major LLM provider — routing, spend, logs, and controls in one place.

New York

130 E 59th St, 17th floor

New York, NY 10022

Wilmington

1201 N. Market Street, Suite 200

Wilmington, DE 19801

Offices