AI agents are changing how applications interact with data, and database security needs to evolve with them. In this July 2026 edition of the Azure Database Platform Security Newsletter, we focus on what this shift means for identity, access, monitoring, and data protection, along with the latest security feature updates and practical resources to help teams strengthen their database security posture.

AI agents are quickly becoming a new layer in how applications interact with data. Unlike traditional applications that follow predictable workflows, agents operate with more autonomy. They can generate queries dynamically, explore data, and act on behalf of users or systems.

This shift changes the database security model.

Agents do not just introduce new risks, they amplify existing ones. Over-privileged access can scale quickly, identity boundaries can become harder to trace, and dynamically generated queries can increase the risk of unintended data exposure or manipulation.

In this new reality, every AI agent effectively becomes a database user with the ability to access, reason over, and act on sensitive data.

Securing data in an agentic world requires a shift in mindset:

• Identity-first security: every agent should have a unique, traceable identity.


• Least privilege by default: access should be tightly scoped to the task at hand.


• Data-aware protection: only the data that is strictly required should be exposed.


• Continuous monitoring: assume issues can occur and detect them early.

Azure SQL provides a broad set of security capabilities, from authentication and fine-grained access control to encryption, masking, auditing, and threat detection. Together, these capabilities enable a defense-in-depth approach for agent-driven scenarios.

Feature highlights 💡

Transparent Data Encryption in Azure SQL Database now supports AES keys (Preview)

Azure SQL Database now supports symmetric AES keys for Transparent Data Encryption (TDE) with customer-managed keys in public preview. This gives organizations another option for protecting data at rest while keeping control of key management.

For teams thinking about long-term cryptographic resilience, this preview is especially relevant. TDE with customer-managed keys has traditionally used asymmetric RSA-based key protectors, while broader industry guidance is increasingly focused on preparing for a post-quantum cryptographic future and adopting cryptographic approaches that are better aligned with that transition.

This update aligns with broader security guidance, including the NSA’s CNSA 2.0 recommendations, which emphasize modern cryptographic planning for a quantum-resistant future. For organizations building crypto agility into their platforms, AES support is a practical step in that direction.

Learn more: Transparent data encryption in Azure SQL Database now supports AES keys (Public Preview)

General Availability: Microsoft Entra server principals for Azure SQL Database

Azure SQL Database has historically lacked support for server-level Microsoft Entra principals, requiring customers to manage identities separately in each database. This has been a migration blocker for enterprise customers coming from SQL Server, where centralized login and access management is standard.

With this release, Azure SQL Database now supports server-level Microsoft Entra principals. Customers can create identities once, apply them consistently across databases, and operate with a unified, enterprise-grade identity and access model aligned with SQL Server and Azure SQL Managed Instance.

Learn more: Generally Available: Microsoft Entra Server Principals and Server Roles for Azure SQL Database.

Cross-tenant customer-managed keys for Azure Database for PostgreSQL Flexible Server (Preview)

This preview allows customers to encrypt data at rest using an Azure Key Vault key that resides in a different Microsoft Entra tenant than the database service. It is especially relevant for ISVs and customers in regulated industries that need clear separation of duties between key ownership and service operation.

Customers retain control over key lifecycle management, including rotation and revocation, while authentication to the customer-owned Key Vault is handled through federated identity with Microsoft Entra ID for cross-tenant access.

Learn more: Data Encryption – Azure Database for PostgreSQL | Microsoft Learn

Defender security posture assessments for Azure Database for PostgreSQL Flexible Server (GA)

Microsoft Defender Cloud Security Posture Management (CSPM) assessments for Azure Database for PostgreSQL Flexible Server are now available in general availability. This capability introduces built-in security assessments that continuously evaluate PostgreSQL server configurations against best practices, helping identify vulnerabilities and misconfigurations with actionable remediation guidance. A core set of assessments is available at launch, with additional coverage planned for future releases. Assessments are automatically available for PostgreSQL servers where Microsoft Defender for Cloud is already enabled, requiring no additional setup or costs.

Learn more: Microsoft Defender for Cloud – Azure Database for PostgreSQL | Microsoft Learn

Best Practices Corner ⚖️

Move to secure-by-default configurations

Modern data platforms should reduce the need for manual hardening by enabling strong defaults out of the box.

• Enforce encrypted connections by default.


• Disable insecure or legacy options where possible.


• Adopt platform defaults that align with compliance standards.

This shifts security from being configuration-driven to posture-driven, reducing the risk of misconfiguration, which remains one of the leading causes of data exposure.

Classify data first, protect it second

You cannot secure what you do not understand. Start by identifying where sensitive data lives, then use that understanding to apply the right protections.

• Use Data Discovery & Classification to identify sensitive data.


• Apply labels and information types consistently.


• Use those labels to drive masking, encryption, and governance policies.

Why this matters: Security controls are only effective when they are data-aware. Classification provides the foundation for applying the right protection to the right data, especially for regulated workloads.

Continuously assess and remediate vulnerabilities

Security posture is not static. It requires continuous validation and timely remediation.

• Run Vulnerability Assessment scans regularly.


• Track deviations from baseline security policies.


• Act on remediation guidance proactively.

Why this matters: Even well-configured systems drift over time. Continuous assessment helps detect misconfigurations early and maintain a strong, compliant security posture without relying on manual reviews.

Blogs and Video Spotlight 🅱️

Over the last three months, we published several posts covering major releases, platform updates, and practical security guidance. These resources provide deeper technical context and implementation guidance for teams working across Azure SQL, Azure Database for PostgreSQL, Azure Database for MySQL, and Azure Cosmos DB.

Azure SQL Retiring The No Minimum TLS Option

Dynamic Data Masking – What it is, What it isn’t, and How to use it effectively | Microsoft Community Hub

TLS Certificate Pinning in PostgreSQL and MySQL: Risks, Rotations, and Best Practices

I’m Starting a New Cosmos DB App. What Security Do I Actually Need? – Azure Cosmos DB Blog

Transparent data encryption in Azure SQL Database now supports AES keys (Public Preview) | Microsoft Community Hub

Which Azure Cosmos DB Role Does My App Need? – Azure Cosmos DB Blog

Generally Available: Microsoft Entra Server Principals and Server Roles for Azure SQL Database | Microsoft Community Hub

Community & Events 👥

The data platform security team will be present at several community and industry events. If you are attending, come and say hello.

Where to meet us next

Microsoft Fabric and SQL Conference, Barcelona, 28 September to 01 October 2026

dataMinds Connect, Mechelen, 12 to 14 October 2026

Techorama, Utrecht, 26 to 28 October 2026

Recent community highlights

SQLBits, Wales, 22 to 25 April 2026

Azure SQL Bangalore User Group Meeting, Bangalore, 13 June

Call to action 📢

As AI agents become a more active layer between applications, users, and data, now is the right time to review how they access your database platform. Treat every agent as a database user: confirm that each one has a unique identity, validate that permissions follow least privilege, and monitor agent-driven activity for unusual access patterns. A focused access review today can help reduce the risk of unintended data exposure as agentic workloads continue to grow.