HIPAA BAA scope for Azure OpenAI models GPT-6.1 Terra and GPT-6.1 Sol

elton Crowder 0 Reputation points
2026-11-17T17:41:32.82+00:00

We're a HIPAA covered entity, a chiropractic clinic, using Azure OpenAI in our own subscription under the Microsoft Product Terms and HIPAA BAA.

  1. Are the GPT-6.1 Terra and GPT-6.1 Sol models, deployed through Azure OpenAI in our subscription, covered under Microsoft's HIPAA BAA, the same as other Azure OpenAI models?
  2. Does coverage depend on the deployment type (Global Standard, Data Zone Standard, or Regional Standard)? If so, which types are covered for PHI?
  3. Is image input (for example, photos of patient insurance cards) covered the same way as text input?
  4. Is there a published list of Azure OpenAI models and features in scope for the BAA that we can keep on file?
Azure OpenAI in Foundry Models
0 comments No comments

1 answer

Sort by: Most helpful
  1. AI answer

    2026-11-17T17:42:53.5566667+00:00

    Do not place PHI into GPT‑6.1 Terra or GPT‑6.1 Sol until your compliance file confirms that Azure OpenAI/Models sold by Azure is an in-scope service and that the selected deployment’s processing geography satisfies your requirements; Microsoft does not publish a separate HIPAA approval for every model, modality, and deployment combination.

    Direct answers

    1. Are GPT‑6.1 Terra and Sol covered like other Azure OpenAI models?

    • Microsoft classifies gpt-6.1-terra and gpt-6.1-sol as Azure OpenAI models sold by Azure, hosted and operated by Microsoft and billed through your Azure subscription—not as partner/community models. Both listed versions are generally available rather than marked preview. ^1^
    • Microsoft’s BAA applies to in-scope Azure services, rather than providing a model-by-model certification. The controlling scope is the Microsoft Product Terms/DPA and the current Azure audit-scope artifact. ^2^ ^3^
    • Therefore, their classification as Azure OpenAI models supports treating them under Azure OpenAI’s service-level terms, but the public material does not provide a model-specific statement saying “GPT‑6.1 Terra/Sol are HIPAA-approved.” Microsoft also emphasizes that a BAA supports—but does not itself establish—your clinic’s compliance. ^2^

    2. Does the deployment type affect BAA coverage?

    The documentation does not identify Global Standard, Data Zone Standard, or Regional Standard as separate HIPAA eligibility classes. It says Azure data-processing and compliance commitments remain applicable to Global and Data Zone deployments. What changes is where inference processing can occur: ^4^

    Deployment type Processing location Practical PHI consideration
    Global Standard Any Azure geography where the model is deployed Do not select it if your risk assessment or contractual obligations require US-only processing.
    Data Zone Standard Within the selected US, EU, or APAC data zone For a US clinic requiring US-only inference processing, use a Foundry resource and Data Zone deployment in the US.
    Standard / geography-based Within the customer-selected Azure geography, potentially across regions in that geography Use when a specific Azure geography is required and the model is available there.

    Both GPT‑6.1 models are documented for Global Standard and Data Zone Standard. Regional/Standard availability must be checked for the exact model, version, and resource region; not every model supports every deployment type. ^5^

    For a US PHI workload, Data Zone Standard in the US is the clearest documented choice when US-only processing is required. This is a data-location choice, not a statement that Global Standard is outside the BAA.

    3. Is image input covered like text input?

    Both GPT‑6.1 models officially support text and image input with text output. ^6^

    Microsoft’s service data-handling documentation addresses prompts and completions containing both text and images under the same Models sold by Azure processing framework. It says prompts and completions are not available to OpenAI or other model providers, are not used to train foundation models without permission, and may be analyzed for content safety and abuse monitoring. ^7^

    However, there is no public model-and-modality BAA matrix explicitly naming “insurance-card image input to GPT‑6.1.” Treat the image as PHI throughout your architecture:

    • Avoid placing it in application, gateway, diagnostic, or exception logs.
    • Review whether the Responses API or another stateful feature persists message history.
    • Apply least-privilege RBAC and network controls.
    • Document retention and deletion behavior.
    • Review abuse monitoring: flagged prompts or images may enter automated or human review processes by default. Eligible managed customers can apply for modified abuse monitoring, but it is not automatic. ^8^

    What to retain for your compliance file

    Keep dated copies of:

    1. Your governing Microsoft licensing agreement, Product Terms, DPA, and incorporated HIPAA BAA.
    2. The HIPAA offering page and current Microsoft Azure Compliance Offerings audit-scope artifact referenced by the cloud-services-in-audit-scope page. ^2^ ^3^
    3. The GPT‑6.1 catalog page showing the precise model IDs, versions, Azure-sold classification, and image capability.
    4. The deployment-type page showing the processing boundary selected.
    5. The service data-privacy and abuse-monitoring pages.
    6. A Microsoft support response tied to your agreement confirming the exact combination: Azure OpenAI + model/version + deployment SKU + US data zone + image inputs + APIs/features used.

    There is currently no single published list that combines every Azure OpenAI model, modality, feature, and deployment type into a model-level HIPAA BAA matrix. The authoritative contractual record is the BAA/Product Terms plus the current service audit-scope documentation.


    References

    1. Foundry Models sold by Azure (azure-openai)
    2. HIPAA (US)
    3. Azure, Dynamics 365, Microsoft 365, and Power Platform services in audit scope
    4. Deployment types for Azure AI Foundry Models
    5. Region availability for Foundry Models sold by Azure (standard)
    6. Azure OpenAI reasoning models
    7. Data, privacy, and security for Models sold by Azure in Microsoft Foundry
    8. Abuse monitoring
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.