Skip to main content
Zero Data Retention (ZDR) means that a provider will not store your data for any period of time.

​
How Blackbox Manages Data Policies

Blackbox works with providers to understand each of their data policies and structures the policy data in a way that gives you control over which providers you want to route to. Note that a provider’s general policy may differ from the specific policy for a given endpoint. Blackbox keeps track of the specific policy for each endpoint, works with providers to keep these policies up to date, and in some cases creates special agreements with providers to ensure data retention or training policies that are more privacy-focused than their default policies.
If Blackbox is not able to establish or ascertain a clear policy for a provider or endpoint, we take a conservative stance and assume that the endpoint both retains and trains on data and mark it as such.

​
Per-Request ZDR Enforcement

You can enforce Zero Data Retention on a per-request basis using the zdr parameter in your API calls.

​
Usage

Include the zdr parameter in your provider preferences: 
curl -X POST https://api.blackbox.ai/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello"}],
    "provider": {
      "zdr": true
    }
  }'
When zdr is set to true, the request will only be routed to endpoints that have a Zero Data Retention policy. When zdr is false or not provided, ZDR enforcement will still apply if enabled in your account.

​
Codex Models and Encrypted Reasoning

OpenAI Codex models (openai/gpt-5.3-codex, openai/gpt-5.2-codex) always operate under Zero Data Retention — store is enforced to false for every request, regardless of what you pass. This means these models cannot use the Responses API in a stateful way; no data is persisted between requests. To support stateless multi-turn workflows under ZDR, Blackbox uses encrypted reasoning items. The include: ["reasoning.encrypted_content"] parameter is always enabled for Codex model requests, even if you do not provide it in your request. This ensures that encrypted reasoning tokens are returned in every response, which you can pass back in future requests just like regular reasoning items. When a request includes encrypted_content, it is decrypted in-memory (never written to disk), used for generating the next response, and then securely discarded. Any new reasoning tokens are immediately encrypted and returned to you, ensuring no intermediate state is ever persisted. This allows you to maintain full reasoning continuity across multi-turn tool calling while remaining fully ZDR-compliant. For complete code examples showing how to use encrypted reasoning with tool calling, see Tool Calling with Encrypted Reasoning.
This applies only to Codex models (gpt-5.3-codex and gpt-5.2-codex). Other models are not affected.

​
Caching

Some endpoints/models provide implicit caching of prompts. This keeps repeated prompt data in an in-memory cache in the provider’s datacenter, so that the repeated part of the prompt does not need to be re-processed. This can lead to considerable cost savings. Blackbox has taken the stance that in-memory caching of prompts is not considered “retaining” data, and we therefore allow endpoints/models with implicit caching to be hit when a ZDR routing policy is in effect.

​
Blackbox’s Retention Policy

Blackbox itself has a ZDR policy; your prompts are not retained.