Add PII masking to OpenAI

Migration shape

1. Store your OpenAI key once as a BYOK credential in Privian. Remove it from your application environment.
2. Create a Privian gateway key. Set it as PRIVIAN_GATEWAY_API_KEY.
3. Change the OpenAI SDK base_url / baseURL to https://api.privian.io/v1.
4. Use a Privian-supported, provider-namespaced model id (e.g. openai/gpt-4o-mini).

Python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_PRIVIAN_GATEWAY_API_KEY",
    base_url="https://api.privian.io/v1",
)

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

print(response.choices[0].message.content)

Node / TypeScript

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);

What changes vs. calling OpenAI directly

• The OpenAI Chat Completions request and response shapes are unchanged.
• Sensitive values in every message are masked before any provider call.
• Model ids are provider-namespaced (openai/<id>, anthropic/<id>, …) and must be enabled in Privian's model catalog.
• Streaming, tool/function calling, JSON mode, and multimodal content parts are not yet supported and return a fail-closed error.

Related

• Protect prompts before GPT
• BYOK
• API Reference