Microsoft Entra: Are You Still the Same Person You Were Five Minutes Ago?

There was a time when protecting a digital identity seemed almost simple. A user had an account. They knew their password. They authenticated themselves, possibly with a second factor, then accessed the resources they were entitled to. Access was granted. And, at least for a while, the issue seemed settled.

That world is gradually disappearing, because identities are no longer immobile. An employee changes roles. A device ceases to be compliant. An external partner completes its mission but keeps its guest account. An application creates its own groups. A perfectly legitimate session just minutes earlier may be confronted with a new risk signal. And now, a new player complicates the equation  even more: the artificial intelligence agent.

The new features presented by Microsoft in November 2026 precisely tell the story of this transformation. Behind passkeys, Continuous Access Evaluation, Global Secure Access, or new governance policies, a single idea is finally emerging:

Access can no longer be considered a final decision. It must be continuously reassessed.

This is probably the real takeaway from the November 2026 edition of What’s new in Microsoft Entra: November 2026, published on November 6.

The time for lasting trust is over

For a long time, we essentially sought to answer one question: “Does this person have the right to access this resource?” , then Zero Trust came to change this logic. We began to wonder:

“Who is this person?”
“Is its authentication strong enough?”
“Is his equipment compliant?”
“From which location does it connect?”
“Is there a particular risk at the time of connection?”

With the current developments of Entra, a new question arises:

“Are the conditions that allowed us to grant this access still valid at this very moment?”

This nuance may seem subtle. Yet it profoundly changes the way we view identity security.

Rights that drift over time

Let’s take a phenomenon that all administrators know. A user joins a project. They are assigned a group, then an app, a Teams space, a SharePoint site, and maybe a few additional permissions. Six months pass.

He changed his role.
The project is evolving.
Some permissions disappear, others are added manually.
And gradually what could be called a Drift of rights.

Microsoft is precisely introducing drift reports in the Access Packages in Public Preview. The goal is to identify situations where the rights actually assigned no longer correspond to what the governance policy provided. These reports are updated daily and require Microsoft Entra ID Governance or Microsoft Entra Suite.

This is a more significant evolution than it seems, because governing identities is no longer just about correctly defining access. It is now necessary to verify that reality remains consistent with the initial intention.

In other words, IAM also enters into a drift detection logic.

Guests: entering is easy, leaving must become so too

This issue is even more evident with external users.

A consultant is invited for three months.
A partner is added to a Teams project.
A temporary application is opened to a service provider.

The project ends. Then everyone moves on to the next project.

But the Guest account, on the other hand, can quietly continue its existence. This is how some organizations accumulate over time a true phantom population of external identities.

Microsoft now offers, in Public Preview, guest lifecycle policies. They can be based on various criteria such as recent activity, invitation redemption status, or sponsor certification. When a guest no longer meets the defined conditions, it becomes possible to automate their deactivation or deletion.

Here we go beyond a simple periodical review. The Access Review allows us to ask the question:

“Does this person still need to have this access?”

The lifecycle policy provides the logical follow-up:

“If that’s no longer the case, let’s really eliminate this access.”

The difference is fundamental. Good governance is not just  about knowing who can come in. It must also  know when someone should leave.

Passwordless authentication takes a new step

For several years, Microsoft has encouraged us to move away from passwords. But the transition was sometimes paradoxical. To benefit from a modern authentication method, the user sometimes had to start their journey with a much less modern method.

In November 2026, Microsoft announced the general availability of the ability to save a passkey or passwordless method as the first MFA method. Deployment is expected to continue until February 2027.

This evolution might seem secondary, but it is not. It allows for a much more natural envision of a journey where the user enters the company directly with a phishing-resistant authentication mechanism.

Instead of a path like: Password → SMS → MFA → Passkey

We can gradually consider:

Identity→ Secure enrollment→ Passwordless → Phishing-resistant authentication

Passwordless is then no longer just the completion of a migration. It can become the starting point.

Windows Hello is becoming even more important

This evolution is reinforced by another new feature:

Windows Hello for Business and macOS Platform SSO can now be used as standalone MFA factors for various scenarios, including Authentication Strengths, additional authentication requests, or certain sign-in frequency controls.

For organizations that have already widely adopted managed devices, the message is interesting. The workstation is no longer just a terminal from which the user logs in. It itself becomes a component of trust. We see an architecture emerging in which:

the user’s identity,
the identity of the device,
the authentication method,
Its compliance status
and the context of access

all participate in the same decision. This is exactly the direction the Zero Trust model has been suggesting for several years.

A legitimate session may cease to be legitimate

This is where Continuous Access Evaluation, or CAE, makes perfect sense. Let’s imagine a perfectly legitimate user. They log in at 8:30 a.m., their authentication is successful, their device is compliant, and Conditional Access authorizes access.

At 10:12 a.m., something changes. The account suddenly presents a high risk, or the equipment used is no longer considered compliant. In a traditional approach, we might be tempted to wait for a token to expire or re-authenticate before making a security decision.

But this approach is becoming insufficient.

Universal CAE can now use user and equipment risk signals. If a device is disabled or becomes non-compliant, or if a high user risk is detected, the Global Secure Access client can trigger a reauthentication within about two minutes. This may be one of the most representative changes to Entra’s current philosophy.

Trust becomes temporal. It existed five minutes ago, it can disappear now And access must be able to follow this evolution.

When identity and network start to blur

This transformation no longer concerns only Entra ID. Microsoft continues with Global Secure Access the alignment between identity control and network control.
In public Preview, Microsoft Entra Internet Access Cloud Firewall now covers Global Secure Access client traffic to enforce more granular rules on outbound network connections.

At the same time, Microsoft is also trying to solve much more everyday problems.

With Prefer Local Network, a user working from home can continue to access certain resources on their local network, such as a printer or router, without having to disable Global Secure Access. And with support for custom HTTP headers in Entra Internet Access, it also becomes possible to consider finer restrictions on certain SaaS services like GitHub or Slack, with TLS inspection required for HTTPS destinations.

We are therefore no longer just in a logic:

“This URL is allowed or prohibited.”

We move on to a more subtle question:

“Can this user, with this identity, from this device and in this context, use this service in this way?”

The historical boundary between IAM and network security is becoming increasingly less clear.

Even hybrid identity continues to evolve

Another announcement will particularly interest those who, like many companies, still have one foot in Active Directory and the other in Entra ID. Entra Cloud Sync can now synchronize Active Directory computer objects to Entra ID to enable Hybrid Join. Device synchronization can be enabled in the Cloud Sync configuration. This change deserves to be monitored.

For a long time, Entra Cloud Sync has evolved in the shadow of the traditional Microsoft Entra Connect Sync. But little by little, its capabilities are expanding. This obviously does not mean that all existing Entra Connect architectures should be immediately abandoned. Rather, it means that the choice made five or ten years ago now deserves to be reconsidered in light of current capabilities.

In hybrid identity, nothing should be considered eternal 😉

And then the AI agents arrive…

That’s probably the part that interests me most in this announcement. Until now, we’d learned to manage users. Then guests. Then applications, workloads, and technical accounts.

Now a new population must be added: AI agents.

Microsoft introduces mechanisms for managing agent identities in Public Preview. It becomes possible to define requirements regarding sponsors, reconfirmation intervals, or rules based on their inactivity.

Microsoft also describes a default inactivity policy available to all tenants without an additional license: after 60 days of inactivity, the owner is notified and, if no new activity is detected, the agent’s identity can be disabled. And this topic will become fundamental.

Because an AI agent is not just a pretty “Copilot” button. As soon as it has an identity and can consult information, call an API, access an application, or act on a resource, old IAM questions come back immediately.

Who created this agent?
Who is responsible?
Why does it still exist?
What data can they access?
Does he have too many privileges?
What happens when its owner leaves the organization?
And above all, who will think of removing it when the project is finished?

It took us years to discover how problematic forgotten service accounts could become. Let’s try not to repeat exactly the same story with AI agents.

Identity is no longer a state, but a movement

In the end, this might be how to read the Entra new releases in November 2026. Individually, some functions seem relatively technical. An improvement to Cloud Sync here. A new Global Secure Access mechanism there. A change in MFA enrollment elsewhere.

But when you take a step back, a coherence appears. Microsoft is gradually building a world where trust is no longer fixed. A user can be legitimate today and no longer need access tomorrow. A device can be compliant this morning and pose a risk this afternoon. A guest may be needed during a project and then become useless.

And an AI agent can be created in minutes, have considerable permissions, and then be forgotten a few months later. Security must therefore learn to evolve at the same pace.

Conclusion

The November 2026 new features could easily be read as an additional list of features added to Microsoft Entra. In my opinion, that would be missing the point. What Microsoft is gradually building is a trust system capable of living with constant change

Trust given yesterday must never become a right acquired for tomorrow.

And with the massive arrival of AI agents in our information systems, this phrase is likely to gain even more importance.

Leave a Reply