Get Started

AI Security Layer · Beta

The privacy-first LLM gateway for secure AI

Mask personal and sensitive data before prompts reach GPT, Claude and other models — with PII masking, prompt security and a privacy-first gateway.

Helping AI startups become enterprise-ready: pass security reviews faster, protect customer data, and shorten enterprise sales cycles.

Get StartedRead the BlueprintEnterprise-Ready AI →
  • BYOK
  • Zero retention
  • Provider agnostic
  • OpenAI-compatible
Request flowmasked → rehydrated

  1. Client Application

    Your app or backend

  2. Privian

    Mask sensitive data

  3. LLM Provider

    GPT · Claude · Gemini

  4. Privian

    Rehydrate safely

  5. Client Application

    Original values restored

Prompt Privacy

The privacy layer between your app and every LLM.

One endpoint. Every major provider. PII masked before it leaves your network, rehydrated on the way back — with BYOK and zero retention by default.

Drop-in compatible with the OpenAI SDK. Swap providers with a single field.

Get startedRead docsReview Blueprint →
Enterprise-ready by default — learn more
privian.ts
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.PRIVIAN_GATEWAY_API_KEY,
  baseURL: "https://api.privian.io/v1",
});

const response = await client.chat.completions.create({
  model: "openai/gpt-4o-mini",
  messages: [
    { role: "user", content: "Hello, my name is Michael" },
  ],
});

console.log(response.choices[0].message.content);

Use it with

OpenAIAnthropicGeminiGroqAzureMistralOllamaand more

Built with design partners shipping AI into regulated industries

See our Trust Center →

PII masking

How Privian masks sensitive data before prompts reach an LLM

The model never sees the original values. Privian detects personal and sensitive data, replaces it with deterministic placeholders, then restores the originals in the response.

1 — Inbound prompt (from your app)

My name is Michael Andersen, email michael@email.com and my SSN is 123-45-6789.

2 — Sent to GPT / Claude (masked)

My name is [PERSON_1], email [EMAIL_1] and my SSN is [ID_1].

3 — Model response (still masked)

Hi [PERSON_1], I've noted the email [EMAIL_1] for follow-up.

4 — Returned to your app (rehydrated)

Hi Michael Andersen, I've noted the email michael@email.com for follow-up.

Placeholders are deterministic within a request, so rehydration is exact. Learn more about PII masking →

How it works

Three moves in the request path

  1. Step 1

    Route prompts through Privian

    POST to one gateway endpoint with a JSON body — prompt, model and an optional stream flag. The gateway handles provider routing using your BYOK credentials.

  2. Step 2

    Mask sensitive data

    PII masking and prompt security happen before the request reaches the model. Detected values are replaced with deterministic placeholders.

  3. Step 3

    Rehydrate responses

    Sensitive values are safely restored in the model response so your application receives a normal, correctly-formed answer.

Capabilities

Built for secure LLM traffic

A focused set of primitives for LLM security, prompt privacy and sensitive data masking — composed into a single request path.

Enterprise readiness

From AI feature to enterprise deployment

The Enterprise-Ready AI path
  1. Stage 1

    AI startup

    Ship an AI feature

  2. Stage 2

    Security review

    Answer the questionnaire

  3. Stage 3

    Enterprise approval

    Pass procurement

  4. Stage 4

    Production deployment

    Scale on BYOK

Who Privian is for

Built for teams shipping AI on sensitive data

Privian fits teams using GPT, Claude and similar models where prompts regularly contain personal or sensitive content.

  • Customer Support AI

    Problem
    Support tickets are full of names, emails and account data.
    How Privian helps
    Privian masks customer PII before the prompt reaches the model.
    Learn more

  • Internal Copilots

    Problem
    Employees paste internal data into external LLMs.
    How Privian helps
    Route copilot traffic through Privian so sensitive content is masked first.
    Learn more

  • AI Features for SaaS

    Problem
    Shipping AI features means sending user data to a provider.
    How Privian helps
    Add a privacy layer between your product and the model — no prompt rewrites.
    Learn more

  • Secure AI Prototyping

    Problem
    Prototyping with real data is fast but risky.
    How Privian helps
    Use the same gateway in dev so experiments never leak real values.
    Learn more

Trust

Built for trust

What Privian does not do is as important as what it does.

Trust CenterEnterprise Trust Package

  • No raw prompt storage

    Inbound prompts pass through memory only.

  • No rehydrated response storage

    Outbound responses are never persisted.

  • BYOK architecture

    Provider credentials are yours, encrypted at rest.

  • Provider agnostic

    OpenAI, Anthropic, Google — no lock-in to one model.

  • Simple JSON gateway

    One endpoint, one JSON body — prompt, model, optional stream.

  • Minimal metadata retention

    Counters and latencies, never payload bodies.

Transparency

Understand exactly how Privian works before you build

Privian is in beta. Architecture, security model and known limitations are documented publicly — so you can evaluate the product on facts, not marketing.

Developer experience

Integrate in minutes

Privian is OpenAI Chat Completions-compatible. Point the OpenAI SDK at https://api.privian.io/v1 and POST model + messages[]. Sensitive values are masked before egress and rehydrated in the response.

View DocsAPI Reference
Example request
curl -sS -X POST "https://api.privian.io/v1/chat/completions" \
  -H "Authorization: Bearer $PRIVIAN_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o-mini",
    "messages": [
      { "role": "user", "content": "Email me at michael@email.com about ticket #4821." }
    ]
  }'

michael@email.com is masked before reaching the model and rehydrated in the response.

Learn

Understand prompt security, PII masking and LLM privacy

Deeper reading on each part of the Privian AI Security Layer.

Get started

Start building enterprise-ready AI

Protect prompts before they reach GPT, Claude and other models — and give procurement teams the trust artifacts they ask for.

Get Started Read the BlueprintSee pricing →