Registry / Azure OpenAI Service
Azure OpenAI Service
OpenAI GPT models served by Microsoft as "Models sold by Azure" within Microsoft Foundry (formerly Azure OpenAI Service / Azure AI Foundry). Microsoft hosts the models in its own Azure environment; customer data does not flow to OpenAI-operated services, and vendor-trust and data-handling commitments are Microsoft's/Azure's.
Watch-outs 4
The cells where this offering is not a clean public yes. This is what to check before you sign.
- SOC 2 Type II: Yes, sales-gated SOC 2 report can be obtained only through the Microsoft Service Trust Portal, which requires sign‑in or sales contact.
- ISO 27001: No, verified The provider’s official ISO 27001 offering page lists the services in scope and Azure OpenAI Service is not among them, so the certification does not cover this offering.
- Retention / ZDR: Partial zero‑data‑retention (modified abuse monitoring) requires a gated request; retention period not explicitly stated
- Residency: Partial Only geography‑level pinning; processing can move between regions in the same geography.
The provider’s own blog confirms SOC 2 coverage and directs users to the Service Trust Portal, a gated portal, so the report is available but not publicly without login.
Azure OpenAI already meets a wide range of industry standards and certifications, including **SOC 2**, ... ensuring ... compliance. The page links to the Service Trust Portal where audit reports are gated.
The provider’s official ISO 27001 offering page lists the services in scope and Azure OpenAI Service is not among them, so the certification does not cover this offering.
The Azure ISO/IEC 27001 certificate covers Azure, Dynamics 365, Power Platform, and select Microsoft 365 cloud services.
The Microsoft blog publicly confirms Azure OpenAI models are covered by the ISO/IEC 42001:2023 certification.
By achieving this certification, Microsoft demonstrates that Azure AI Foundry Models, including Azure OpenAI models, and Microsoft Security Copilot prioritize responsible innovation and are validated by an independent third party.
Microsoft's Service Trust Portal publicly hosts an AI Resources page that includes Azure OpenAI, confirming a maintained trust center.
Resources describing the approach to Compliance, Security and Privacy in the Artificial Intelligence (AI) solutions such as Copilot and Azure Open AI
Microsoft's authoritative Azure HIPAA offering page confirms the BAA is publicly available by default through the DPA/Product Terms with no gate, so yes_public is correct; corrected the source from a community Q&A page to Microsoft's own compliance page.
The Microsoft Product Terms references the Microsoft Products and Services Data Protection Addendum (DPA), which states that "execution of customer's volume licensing agreement includes execution of the HIPAA Business Associate Agreement".
Replaces the third-party shl.com copy with Microsoft's own pages: the DPA is publicly downloadable on microsoft.com/licensing, Microsoft's white paper states it "released an updated version of the Microsoft Products and Services Data Protection Addendum (DPA) implementing the new [2021] SCC" (Micros
The DPA is an addendum to the Product Terms site (and formerly OST). The current and archived editions of the DPA are available for download.
The Azure OpenAI data‑privacy page publicly states that prompts and completions are not used to train foundation models without explicit customer permission, confirming a public commitment.
- are NOT used to train any generative AI foundation models without your permission or instruction.
Docs confirm data retention is described (human‑review storage) and that zero‑data‑retention is only available via a sales‑gated modified abuse monitoring request, so the claim is only partially satisfied.
Authorized Microsoft employees may assess content flagged... This human review abuse monitoring process won't take place if the customer has been approved for modified abuse monitoring. ... Customers may apply to modify abuse monitoring by completing this form.
Microsoft’s official data‑privacy page confirms processing is limited to the chosen geography but not to a specific region, matching the rubric’s definition of partial.
Prompts and responses are processed within the customer‑specified geography (unless you are using a Global or DataZone deployment type), but may be processed between regions within the geography for operational purposes (including performance and capacity management).
The EC’s official GPAI Code of Practice signatory list, hosted on the European Commission site, includes OpenAI, confirming the model developer is a public signatory.
- OpenAI
The Art. 53 training-content summary is a developer obligation, and per the knowledge base serving-platform rows (e.g. Azure OpenAI Service) are not_applicable, even though developer OpenAI has published its summary.
In accordance with our obligations under Article 53(1)(d) of the EU AI Act, OpenAI publishes summaries about the content used to train our general-purpose AI models:
Spotted an error? Submit a correction with evidence, corrections with a primary source are folded in and credited in the changelog.