Blog Post

Azure Storage Blog
6 MIN READ

Faster, Secure Reads for AI Workloads: Session-Based Authentication Now in Preview for Azure Blob

jeffpatt's avatar
jeffpatt
Icon for Microsoft rankMicrosoft
Nov 17, 2026

Data continues to grow exponentially on Azure Blob and Azure Data Lake Storage for a wide range of workloads: AI training and inference, agentic AI, analytics, high-performance computing, and backup and recovery. Azure Blob and Azure Data Lake Storage are the foundations of these exabyte-scale data estates, and agentic AI is now fundamentally changing how that data is accessed and used.

Azure Storage has always offered granular access control and strong auditing for access to the data. Every request can be authenticated with a Microsoft Entra ID identity, authorized with Azure role-based access control (RBAC) and attribute-based access control (ABAC) conditions, and recorded in Azure Storage diagnostic logs. The same controls apply to AI agents.

What is changing is scale. AI and agentic workloads issue large volumes of read requests, often across many objects in the same container, resulting in increased latency for authentication and authorization. To meet those demands, we are introducing a new authentication method that reduces per-request latency while preserving the security controls organizations already rely on.

Today, we are excited to announce that session-based authentication is now in public preview for Azure Blob and Azure Data Lake Storage, in all public Azure regions.

With session-based authentication, an agent or application authenticates once with Microsoft Entra ID, establishes a short-lived session for a container, and uses that session for subsequent reads. It brings three benefits:

  • Improved performance. Identity and permissions are verified once when the session is created. Subsequent read requests to the container use that session which reduces per-request authentication and authorization latency.
  • The same least privileged access, now short-lived. Sessions use your organization’s existing permission model for least privileged access, Azure RBAC and Azure ABAC. Each session is also scoped to one container, expires after five minutes, is bound to the client that created it, and requires every request to be signed.
  • Full traceability. Azure Storage diagnostic logs currently record the Microsoft Entra ID identity behind each request, and the same is true with sessions. With sessions, every request remains tied to the Microsoft Entra ID identity that created the session, so security teams keep visibility into which identity, human or agent, accessed what data, and when.

With the Azure Storage SDK, enabling session-based authentication is a simple client configuration option, with no changes to your application logic.

Whether you are grounding AI in your company’s knowledge, giving agents access to your data, or feeding a training pipeline, session-based authentication helps you read data faster while preserving security. Read on to learn how sessions work, which workloads benefit most, and how to get started.

How session-based authentication works

When you call the Create Session REST API, Azure Storage verifies your Microsoft Entra ID identity, checks RBAC permissions and ABAC conditions, starts a five-minute session, and returns session credentials. The session credentials consist of two parts:

  • A session token, which identifies the session.
  • A session key, which your client uses to sign each request made within the session.

For subsequent requests to the same container, your client sends the session token together with a signature computed with the session key, instead of a Microsoft Entra ID token. Because identity and permissions were already verified when the session was created, per-request authentication and authorization latency is reduced. Sessions are also bound to the network context of the client that created them, which helps prevent a captured token from being replayed by another client and every request is logged to the diagnostic logs.

For this preview, session tokens are supported only for the Get Blob operation; all other operations require a Microsoft Entra ID access token. The Azure Storage SDKs handle this automatically. To learn more, see the Session Based Authentication documentation.

Workloads that benefit from using session-based authentication

Session-based authentication is a good fit for read-heavy, latency-sensitive workloads that make many requests to the same container using the same identity. These workloads benefit from higher throughput and transaction rates, with lower latency. Examples include:

  • AI training and inference: Training pipelines repeatedly read dataset shards, images, and other inputs, while inference services load model artifacts and supporting data.
  • Inferencing agent: Accelerates data insights to support faster business decisions.
  • AI ops log-investigation agent: Quickly detects anomalies and raises alerts to speed remediation.
  • Analytics: Query engines and processing jobs read many files or portions of files in parallel to filter, aggregate, and transform data.
  • High-performance computing: Simulations and scientific applications read shared input datasets across many workers, creating concentrated demand for fast data access.

Sessions and agent delegation

In an agentic workflow, a user asks an agent to complete a task, and the agent reads data on that user's behalf. Session-based authentication keeps that delegation tightly bounded:

  • Access is inherited, never expanded. The agent creates the session with the user's Microsoft Entra ID identity or with its own agent identity. Azure Storage checks RBAC and ABAC before it creates a session. The agent can read only what that identity is already allowed to read, even if a prompt injection tries to steer it toward other data.
  • Access is short-lived and scoped to the task. Sessions expire after five minutes and are scoped to one container. Reaching another container requires creating a new session and a new authorization check.
  • Access can't be handed off or replayed. Sessions are bound to the network context of the client that created them, and every request is signed with the session key. A token captured from an agent can't be reused by another client.
  • Access stays traceable. Every read in the session is logged against the Entra ID identity that created the session.

Turbo tier – Better Together

To further improve performance for read-heavy AI and data intensive workloads, use Azure Blob Storage Turbo tier which is now in public preview. Turbo tier leverages session-based authentication and other improvements to deliver high throughput and low latency for applications that need higher storage performance. Less performance-sensitive data can stay on cost-efficient HDD-backed access tiers. Smart placement automatically moves eligible data as workload access patterns change, while customers keep a single storage account, namespace, and Blob API surface. To learn more about Turbo tier and the performance benefits, see the Azure Blob Storage Turbo tier blog.

Pricing and Availability

Creating a session is billed as a write transaction. Read requests made with a session are billed at the standard transaction rates for your storage account type. To learn more, please see Azure Blob Storage pricing.

Session-based authentication is in public preview in all public regions. This preview release supports session-based authentication with Microsoft Entra ID at the Blob container level and applies to the Get Blob operation, and is available via REST API and SDKs.

Getting Started

To get started, you need to register for the preview by running the following Azure PowerShell commands:

Connect-AzAccount -SubscriptionId <ID>
Register-AzProviderFeature -FeatureName SessionRequest -ProviderNamespace Microsoft.Storage

To check your registration state, run the following Azure PowerShell commands:

Connect-AzAccount -SubscriptionId <ID>
Get-AzProviderFeature -FeatureName SessionRequest -ProviderNamespace Microsoft.Storage

Once your registration is approved, the RegistrationState will show as Registered.

You can also use Azure CLI to register for the preview. To learn more, see the Session Based Authentication documentation.

Using sessions with the Azure Storage SDK

The Azure Storage SDK offers the simplest way to use session-based authentication. Enable session mode when you initialize your Azure Blob client; the rest of your application logic stays unchanged.

Python SDK sample

The only change is enabling session-based authentication with use_session=True.

with DefaultAzureCredential() as credential:
    with BlobServiceClient(
        account_url=account_url,
        credential=credential,
        use_session=True,
    )

For SDK examples, see the Session Based Authentication documentation.

When session mode is enabled, the SDK manages the full session lifecycle automatically:

  • Creates a session and caches its credentials.
  • Automatically signs every request with the session key.
  • Uses the session token for Get Blob operations and Microsoft Entra ID access token for other operations.
  • Renews the session before expiration without affecting data access, so your app never needs to track session lifetimes.

Your existing code continues to read blobs with the same methods, Microsoft Entra ID credentials, and RBAC role assignments, while benefiting from lower-latency reads.

More Information

To learn more, see the Session Based Authentication documentation.

Help and Support

If you have questions or feedback, please reach out to us at blobsessionauth@microsoft.com. If you need technical help, get answers from community experts in Microsoft Q&A or create a support request:

  • For Service type, select Blob Storage. 
  • For Problem type, select Authentication and Authorization.
  • For Problem subtype, select Issues with Session-based authentication.
Updated Nov 17, 2026
Version 1.0
No CommentsBe the first to comment