Skip to main content

A major concern for enterprise AI adoption is data privacy. When you send proprietary code, customer support tickets, or financial data to an LLM provider (like OpenAI or Anthropic), you need a guarantee that they aren’t saving that data to train their future models. LLM Router itself operates on a strict Zero Data Retention (ZDR) architecture—we never store your prompts or completions. However, the upstream providers you route to might still log your data unless explicitly configured otherwise. To solve this, LLM Router offers the zdr configuration flag.

How the ZDR Flag Works

When you enable zdr: true in your request, LLM Router will only route your prompt to providers and models that have a strict, contractual Zero Data Retention policy or allow us to send API headers (like X-Opt-Out) that legally prevent them from storing or training on your data. If your standard routing configuration (e.g., your order or tags) attempts to use a provider that does not support ZDR, LLM Router will automatically block that route and fall back to a compliant provider.

Configuration

You configure this behavior inside the gateway object.
TypeScript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.llmrouter.app/v1",
  apiKey: process.env.LLM_ROUTER_API_KEY,
});

async function main() {
  const response = await client.chat.completions.create({
    model: "anthropic/claude-3-5-sonnet", // Or leave blank if using Tags
    messages: [
      {
        role: "user",
        content: "Analyze this highly confidential Q3 earnings report...",
      },
    ],

    // @ts-expect-error - Custom LLM Router extension
    gateway: {
      zdr: true, // Force routing exclusively to Zero Data Retention providers
    },
  });

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

Configuration Properties

PropertyTypeDefaultDescription
zdrbooleanfalseWhen true, LLM Router will actively filter your fallback order and tag routing to only allow providers that guarantee Zero Data Retention and no model training on API inputs.

Redaction Synergy

For maximum enterprise security, combine zdr: true with our Data Redaction (gateway.redact) features. This ensures that even if a compliant provider experiences a breach, your most sensitive PII was never transmitted to them in the first place.