Azure OpenAI Service — privacy & data policy

Provider: Microsoft. PromptPrivacy score: 88/100. Last verified November 01, 2026.

What happens to your data?

Your prompts and completions are processed in your designated Azure environment and are never used to train foundation models or shared with model providers like OpenAI. By default, inputs and outputs may be processed by automated filters and temporarily stored for abuse monitoring, unless your organization is approved for modified abuse monitoring. Uploaded files, stateful entities, and fine-tuned models reside encrypted in your Azure tenant and can be deleted by you at any time.

Does it train on your data?

Does not train on your content. Prompts, completions, embeddings, and fine-tuning datasets are not used to train, retrain, or improve any generative AI foundation models or services by Microsoft, OpenAI, or other providers without explicit instruction or permission.

Can you opt out?

Not applicable. Azure OpenAI Service does not train foundation models on customer prompts, completions, embeddings, or training data by default.

Data retention

Base models are stateless during inferencing, meaning no prompts or completions are stored in the model. Stateful features (such as Assistants API threads, batch data, and custom fine-tuned models) retain customer data until the customer chooses to delete it. For abuse monitoring, flagged prompts and completions may be stored in a logically separated store for automated and human review, though the specific retention duration is not stated in this document. Customers approved for modified abuse monitoring have prompt logging disabled.

Encryption & security

Data stored at rest is encrypted by default using AES-256 encryption with Microsoft-managed keys, with the option to configure customer-managed keys (CMK). Abuse monitoring data stores are logically separated by customer resource and accessible only to authorized Microsoft personnel using Secure Access Workstations (SAWs) and Just-In-Time (JIT) access controls. In-transit encryption is not specified in this document.

Is it safe for work?

Yes. Azure OpenAI Service is an enterprise-oriented cloud offering governed by Microsoft's Data Protection Addendum (DPA). It provides tenant data isolation, AES-256 encryption at rest, customer-managed key support, and a contractual guarantee that customer data is not used for model training or shared with external parties.

Source policy: https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy