While the Microsoft Entra “What’s New” page hasn’t been updated with a July 2026 section yet, the Microsoft Entra Blog on Tech Community has been active with three significant announcements this month. Each one addresses a different facet of the modern identity and security landscape — AI agent governance, privileged access control, and data protection in the age of AI.
Here’s what shipped, why it matters, and what your organization should do about it.
1. Governing AI Agent Identities with Entra Agent ID
Published: July 7, 2026 Status: Generally available (Agent ID platform) with rolling governance enhancements
Microsoft Entra Agent ID, which reached general availability in April 2026, is Microsoft’s identity and access platform for AI agents. The July blog post from chiragdayani focuses on a theme that’s becoming central to enterprise security: treating AI agents with the same identity governance rigor as human employees.
The Problem This Solves
AI agents are proliferating across enterprise environments at an unprecedented rate. Built on platforms like Microsoft Copilot Studio, Azure AI Foundry, and Security Copilot — not to mention third-party and custom solutions — these agents can access data, call APIs, and take actions on behalf of users. Without consistent identity controls, they become “shadow AI” — operating without oversight, accountability, or auditability.
Key Capabilities
Agent Registry provides a unified, extensible directory of all agent identities across your tenant. You get a complete inventory of your agent fleet, visible in the Entra admin center and integrated with the Microsoft 365 admin center.
Agent Identity Blueprints are reusable templates that define permissions, roles, and governance policies for classes of agents. Approve a blueprint once, and every agent created from it inherits those settings. If something goes wrong, disabling the blueprint instantly shuts down all associated agents — a critical containment capability.
Lifecycle Management ensures that agent identities always have a delegated human sponsor accountable for their access and lifecycle. When a sponsor leaves the organization, sponsorship automatically transfers to their manager through Lifecycle Workflows. Co-sponsors and managers receive notifications, reducing manual oversight burden.
Conditional Access for Agents extends Zero Trust protections to agent user accounts. Administrators can target agents with Custom Security Attributes, apply policies based on Agent Risk, require compliant devices (including Windows 365 for Agents), and enforce device platform and network conditions.
Access Governance through Entitlement Management provides access packages that ensure agent access is intentional, auditable, and time-bound — the same governance model used for human identities.
Why This Matters
The message is straightforward: AI agents in your environment need the same identity discipline as your human workforce. Agent ID makes that possible using the tools your identity team already knows — Conditional Access, identity governance, lifecycle workflows, and PIM. If you’re already managing identities in Entra, you have the skills to manage agents.
Getting Started
- Navigate to Agent ID in the Entra admin center (below Users, Groups, and Devices in the left nav)
- Review the agent inventory in your tenant
- Define agent identity blueprints for common agent types
- Configure Conditional Access policies targeting agent identities
- Set up Lifecycle Workflows for sponsor management
2. PIM Custom Extensions: Business Logic in Role Activation Workflows
Published: July 1, 2026 Author: Kaitlin Murphy Status: Public Preview
Privileged Identity Management (PIM) in Microsoft Entra ID now supports custom extensions — a capability that lets organizations integrate external business logic directly into role activation workflows. This is one of those features that security and IT teams have been asking for since PIM was first introduced.
The Problem This Solves
Privileged access decisions often depend on context that lives outside of PIM. A change ticket needs to be valid. An employee’s HR status needs to be active. The user needs to be on the on-call schedule. Compliance checks need to pass. Previously, these checks happened manually — if they happened at all — creating gaps that attackers could exploit and audit findings that were hard to resolve.
How It Works
With custom extensions, PIM calls your organization’s REST API during the role activation process:
- User requests role activation in PIM
- PIM sends a structured request to your custom extension API, including principalId, roleDefinitionId, justification, ticketInfo, and scheduleInfo
- Your API evaluates business rules — validates the change ticket, checks HR status, verifies on-call schedules, runs compliance checks
- Your API returns a decision — Approved, AutoApproved, or Denied — along with a reason
- PIM enforces the decision automatically and logs the full interaction for audit
The extension is invoked synchronously during the activation workflow (pre-approval stage), meaning decisions happen in real-time without delays.
Supported Scope
In this preview, custom extensions support:
- PIM for Groups
- PIM for Microsoft Entra roles
- PIM for Azure resources
Setup Steps
- Create a custom extension API — a REST API (HTTP POST) that implements your business logic
- Secure the API with Microsoft Entra ID — register an app and implement token validation
- Onboard the extension in PIM — use Microsoft Graph API to create the custom extension object
- Link the extension to role settings — enable “Require pre-approval custom extension” in PIM role settings
- Activate and validate — test the end-to-end flow by activating a role
Why This Matters
This feature bridges the gap between PIM’s identity-based controls and the business context that determines whether privileged access should be granted. For organizations with mature ITSM processes (ServiceNow, Jira, etc.), this means PIM activations can now validate against change management systems in real-time.
For auditors, the value is clear: instead of seeing “user X activated role Y,” audit logs show “user X activated role Y under business conditions Z, validated against ticket T, with approval from manager M.” Every interaction includes an evaluationId, evaluationOutcome, and reason for end-to-end traceability.
Licensing
- Microsoft Entra ID Premium P2 (required for PIM)
- Power Automate or Azure Logic Apps for workflow integration
- ITSM system licenses for external integrations
3. Protecting Sensitive Data in Motion with Purview and Entra
Published: July 1, 2026 Author: SuleTatar Status: Public Preview
Microsoft announced the public preview of network-layer data security powered by the integration of Microsoft Purview and Microsoft Entra. This extends data detection and identity-aware enforcement to the network layer, helping protect sensitive data as it moves across SaaS apps and AI platforms in real-time.
The Problem This Solves
Data doesn’t stay still anymore. It moves through browser sessions, SaaS applications, generative AI prompts, and cloud repositories — often beyond the visibility of traditional endpoint and at-rest controls. Employees paste sensitive data into consumer AI tools. They upload confidential files to unmanaged cloud storage. They share proprietary information through SaaS apps that IT hasn’t sanctioned.
Traditional Data Loss Prevention (DLP) operates at endpoints and within managed applications. But when data moves through the network to unmanaged SaaS and AI apps, those controls can’t see it — let alone block it.
Key Capabilities
Network-layer detection identifies how sensitive data is shared to:
- Shadow AI tools (unmanaged consumer AI apps like ChatGPT, Claude, Gemini)
- Unmanaged SaaS applications
- Personal cloud repositories
Real-time enforcement blocks or limits data exposure based on three dimensions:
- Data sensitivity — Purview sensitivity labels and DLP policies classify the data
- User identity — Entra identity context determines who is accessing what
- User behavior — Insider Risk Management signals track how users interact with sensitive data over time
Unified investigation correlates signals across:
- Microsoft Purview (data classification, DLP, insider risk)
- Microsoft Entra (identity, access, Conditional Access)
- Microsoft Defender (threat detection, investigation, response)
Coverage areas include browser sessions, SaaS app usage, AI interactions (including prompts and responses), and data in motion across the network.
How It Works
The integration combines:
- Microsoft Purview data classification, DLP policies, and insider risk detection
- Microsoft Entra identity-aware enforcement at the network layer through Global Secure Access (GSA)
- Microsoft Defender for threat correlation and investigation
When a user attempts to share sensitive data to an unmanaged SaaS or AI app, the system evaluates the request against DLP policies, sensitivity labels, user identity context, and behavioral signals. If the data is classified as sensitive and the destination is risky, the request can be blocked or logged with full context.
Why This Matters
This is Microsoft’s answer to one of the most pressing security challenges of the AI era: how do you let employees use AI tools productively while preventing sensitive data from leaking through prompts and uploads?
The approach is significant because it follows the data rather than blocking specific apps. Instead of maintaining a growing blocklist of AI tools (which changes daily), the system evaluates data sensitivity and user context in real-time. If the data is sensitive, it’s blocked regardless of destination. If it’s not, the user can work without friction.
Licensing
- Microsoft 365 E7 — Network data security capabilities are included
- Alternative path: Microsoft Purview ME5 (or equivalent) AND Microsoft Entra Internet Access (or equivalent) licenses
- Microsoft Purview Information Protection / DLP for data classification
- Microsoft Entra ID Premium P1 for Conditional Access
Also Worth Noting: July 2026 Enforcement Milestones
Several previously announced changes reached enforcement milestones in July 2026:
SSPR Registration Campaign (July 6, 2026)
Microsoft automatically launched a registration campaign prompting affected administrators and end users to register authentication methods ahead of the September 7, 2026 enforcement date. After September 7, Self-Service Password Reset will only accept explicitly registered authentication methods — directory-sourced phone numbers and email addresses that were never formally registered will no longer be accepted.
Action required: Ensure users have at least one registered authentication method before September 7 to avoid disruptions.
Conditional Access During Credential Registration (July 6-13, 2026)
Entra ID Conditional Access policies scoped to the “Register security information” user action are now being evaluated during credential registration for Windows Hello for Business and macOS Platform SSO. This ensures registration policies apply consistently across all registration flows. Users must satisfy policy controls (MFA, network restrictions, device compliance) before completing registration.
Action required: If you have CA policies targeting security info registration, test them in report-only mode to verify they work as expected across all registration flows.
Security Defaults Block Device Code Flow (July 1, 2026)
All new Microsoft Entra tenants now block device code flow as part of security defaults. This is a security hardening measure — device code flow is commonly targeted by phishing attacks.
Action required: If you have applications or devices that depend on device code flow, you’ll need to either migrate them to a more secure authentication method or configure Conditional Access to explicitly allow device code flow where needed.
Entra ID Pricing Changes (July 1, 2026)
Commercial pricing for Microsoft 365 SKUs was adjusted effective July 1, 2026. Entra ID P1 increased from $6.00 to $7.00 per user/month. This is part of a broader commercial pricing update across the Microsoft 365 portfolio, coinciding with the addition of new AI capabilities (including Copilot) to many SKUs.
The Bigger Picture: Securing the AI Era
These July 2026 announcements reflect a clear strategic direction from Microsoft: the identity, access, and data protection controls that organizations have built for human identities and managed applications need to extend to AI agents, unmanaged SaaS, and network-layer data flows.
Agent ID governance answers the question: “How do we control AI agents in our environment?” PIM custom extensions answers: “How do we enforce business context on privileged access?” Purview + Entra network protection answers: “How do we prevent sensitive data from leaking through AI and SaaS?”
Together, they form a more complete security model for the agentic AI era — one where identity, access, and data protection are unified rather than fragmented across different tools and teams.
Microsoft is also hosting a three-part webinar series starting July 21, 2026 — “Securing Data and Access in the Era of AI” — covering these themes in depth. The July 23 session specifically addresses AI agent security, including local agents developed on tools like GitHub Copilot CLI, Claude Code, and OpenAI Codex.
What You Should Do Now
Audit your AI agent inventory — Check the Agent ID section in the Entra admin center to see what agents are already in your tenant. You may be surprised.
Evaluate PIM custom extensions — If you have an ITSM system (ServiceNow, Jira, etc.), this preview is worth testing. Start with a non-production role to validate the integration.
Review your DLP strategy — If sensitive data protection is a priority (and it should be), evaluate the Purview + Entra network-layer integration. The preview is available now.
Prepare for SSPR enforcement — The September 7, 2026 deadline is firm. Run a report on registered authentication methods and target users who haven’t registered yet.
Update your security documentation — These changes represent meaningful shifts in how identity and data protection work in the Microsoft cloud. Make sure your team’s documentation reflects them.
Kevin Kaminski is the owner of Big Hat Group Inc., a Microsoft partner focused on AI and cloud technologies. Follow him at https://x.com/kkaminsk for more Microsoft Entra insights.