Data privacy in Azure AI foundry

Question

On the page:
https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy?view=foundry-classic&tabs=azure-portal#how-does-the-azure-openai-service-process-data
it says
Your prompts (inputs) and completions (outputs), your embeddings, and your training data:

• are NOT available to other customers.
• are NOT available to OpenAI or other Azure Direct Model providers.
• are NOT used by Azure Direct Model providers to improve their models or services.
• are NOT used to train any generative AI foundation models without your permission or instruction.
• Customer Data, Prompts, and Completions are NOT used to improve Microsoft or third-party products or services without your explicit permission or instruction.

Your fine-tuned Azure Direct Models are available exclusively for your use.

Foundry is an Azure service; Microsoft hosts the Azure Direct Models in Microsoft's Azure environment and Azure Direct Models do NOT interact with any services operated by Azure Direct Model providers, for example, OpenAI (e.g. ChatGPT, or the OpenAI API).

However, when using the 'ask learn' feature it told me that:

The general terms and conditions of Azure do allow Microsoft to change its policies, including those related to Azure Foundry and subprocessors, without prior notification to the customer. Specifically, Microsoft reserves the right to modify the terms at any time, with or without notice, and continued use of the services following such changes will be deemed acceptance of the modifications. 1 Furthermore, Azure Direct Models are governed by the terms applicable to Microsoft Azure Services, which include provisions for changes in service terms.

The way I understand this, is that at a future point, Microsoft might use customer data for model training and it might send customer data to open ai or other 3rd party providers without notifying its customers.
If this is the case, how can we still trust Microsoft Azure with sensitive data? I am mostly concerned with data that contains intellectual property, not personal data.

Answer 1

Your reading highlights a difference between contractual assurances about data use today and the legal reality that large cloud providers reserve the right to change terms in the future. Both statements can be true at the same time, and understanding how they coexist is key to assessing trust.

First, the statements on the Azure AI Foundry page are binding representations of current service behavior. They are not marketing language; they describe how the service is architected and contractually operated today. In particular, Azure Direct Models running inside Microsoft’s Azure environment are explicitly designed to be isolated from OpenAI-operated services, and Microsoft contractually commits that prompts, completions, embeddings, and training data are not used for training foundation models or shared with third parties without permission. Violating these assurances would expose Microsoft to breach-of-contract claims, regulatory action, and reputational damage, especially from enterprise and government customers.

Second, the “terms may change without notice” language is standard boilerplate across nearly all enterprise cloud agreements, not something unique to AI or Azure. It exists because providers must retain legal flexibility to evolve services, comply with new laws, or deprecate features. Importantly, this clause does not mean Microsoft can retroactively do anything it wants with your data while you continue using the service unchanged. Material changes that affect data ownership, confidentiality, or permitted use would typically trigger updated Data Protection Addendums (DPAs), product-specific terms, or regulatory disclosures, and large customers often have contractual rights to terminate or renegotiate if such changes are unacceptable.

Third, from a practical governance standpoint, Microsoft’s entire cloud business—especially Azure’s role in regulated industries—depends on predictability and enforceability, not just trust. Many Azure customers operate under export controls, trade secret laws, defense regulations, and IP protection regimes. If Microsoft were to begin training models on customer IP or forwarding data to OpenAI or other third parties without explicit opt-in, it would immediately jeopardize major contracts, certifications, and government approvals. That risk vastly outweighs any potential benefit from quietly harvesting customer data.

That said, your concern is valid in principle: no cloud provider can offer an absolute, perpetual guarantee independent of contracts and law. Trust in Azure (or AWS, or Google Cloud) is not based on goodwill alone, but on a layered system of contractual commitments, auditability, compliance regimes, customer leverage, and the ability to exit. For organizations with especially sensitive intellectual property, this is why best practice includes measures such as private networking, customer-managed keys, encryption-at-rest and in-use controls, restricted data scopes, and explicit contractual addenda that lock in data-use restrictions beyond default terms.

---

If the above response helps answer your question, remember to "Accept Answer" so that others in the community facing similar issues can easily find the solution. Your contribution is highly appreciated.

hth

Marcin