Provider Logging & Data Retention

Each AI provider on ModelStream has its own data handling policies for logging and retention. This page explains how to control which providers can access your data.

Provider Data Logging Overview

Last Updated: 21 April 2026

While ModelStream enforces a strict zero-retention passthrough policy, we are an aggregation platform that routes your requests to third-party model providers (e.g., OpenAI, Google, Anthropic).

Each provider maintains their own distinct data logging and retention policies that operate independently of ModelStream.

Standard Provider Policies (Non-Zero Data Retention)

By default, the majority of API providers adhere to a 30-day retention period. During this window, your inference data (prompts and completions) is securely logged on their servers.

• Purpose: This data is strictly utilized by providers to monitor for abuse, enforce their terms of service, and ensure platform safety.
• Training Restriction: Data retained under these standard API agreements is not used by providers to train, fine-tune, or improve their foundational models.

Zero Data Retention Exceptions

A select number of providers offer true Zero Data Retention (ZDR) capabilities on their models. When you route a request to a ZDR-enabled model via ModelStream, your data is processed and immediately purged from the provider’s systems, bypassing their standard 30-day safety logging.

This offers the highest level of privacy and is ideal for highly sensitive or regulated workloads.

How to verify a model's logging status

You can verify the specific data logging policy for any model in our ecosystem by checking the "Provider Data Logging" attribute on the model's dedicated details page within the ModelStream directory.