Cut-paper illustration of a laptop, a brass key, a smartphone, and a directory book with user icons beneath a teal shield with a checkmark

On November 5, Microsoft published a CVE with the highest score the scale allows. On the same page, in bold, it said the vulnerability requires no customer action to resolve.

Both sentences are true, and the gap between them is the whole story. A 10.0 that needs no patch sounds like a contradiction. It is not. It is what a cloud service flaw looks like when the vendor finds it, fixes it on their side, and then tells you about it.

What the flaw is

CVE-2026-109469 is an elevation of privilege flaw in Microsoft Entra Device Registration. The weakness is CWE-639, authorization bypass through a user-controlled key. In plain terms, an attacker on the network, with no account and no help from a victim, could abuse a key the user controls to get the service to grant more privilege than it should.

Microsoft's CVSS 3.1 base score is 10.0, vector AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H. Network reachable, low complexity, no privileges, no user interaction, changed scope, full impact on confidentiality, integrity and availability. That is the maximum the scoring system produces.

Three qualifiers sit next to it, and they matter more than the number. The temporal score is 8.7, because exploit code maturity is unproven while the fix is official and the report is confirmed. Microsoft rates exploitation less likely. It says the flaw was not publicly disclosed and not exploited. Checked November 10, the CVE is not in CISA's Known Exploited Vulnerabilities catalog.

The advisory was published November 5, 2026, revision 1.0, and credits Thomas Knudson as the reporter. It carries the exclusively hosted service tag. That tag is the key to the whole advisory.

There is nothing to install

An exclusively hosted service flaw lives in Microsoft's cloud, not in software you run. There is no KB article, no build number, no update to approve. Microsoft's FAQ answers the obvious question directly: the vulnerability has already been fully mitigated by Microsoft, and there is no action for users of the service to take. The CVE exists so customers can assess impact and decide whether follow up is needed in their own environment. It is transparency, published after the fix.

That is a strange feeling if you run a fleet. Every other 10.0 in your career came with a patch window and a verification step. This one arrives already closed, and that is exactly why it is worth a careful read instead of a quick one.

Why device registration is worth a 10.0

Device Registration is the Entra service that records devices in your directory. A registered device is not just an entry in a list. It is the basis for the Primary Refresh Token that gives a Windows device single sign on, and it feeds the device state your Conditional Access policies evaluate. When a policy says only compliant or joined devices, it is trusting what this service recorded.

Microsoft published almost no technical detail about the flaw itself. No endpoint, no request format, no description of the key or the boundary it crossed. That is normal for a cloud service CVE, and it sets a limit on what anyone outside Microsoft can say about the mechanics. What is public is the weakness class and the component, and the component is enough to take seriously.

In February 2025, Microsoft documented a campaign by a group it calls Storm-2372. Attackers sent device codes disguised as Teams meeting IDs, victims typed them into Microsoft's real device login page and passed their real MFA, and the attackers used the resulting tokens to register their own devices in Entra ID and obtain Primary Refresh Tokens. That technique was not a product flaw. It abused a legitimate sign in flow. But it showed the prize clearly: a device the directory trusts is a durable foothold, and it satisfies the device conditions that Conditional Access checks.

A 10.0 in the same service is a different kind of event, but it points at the same trust. The directory's record of which devices belong to you is load bearing. When the service that keeps that record has an authorization bypass, the score reflects what is at stake, not just how the bug works.

The score measures the flaw, not your exposure

It is easy to read 10.0 as drop everything. Microsoft's own assessment says the realistic picture is calmer: exploitation less likely, no public exploit code, no exploitation observed. The base score answers how bad the flaw is if someone uses it. The rest of the advisory answers how likely that is. Both numbers belong in the same sentence, or the first one misleads you.

The part the advisory cannot answer is the one that will bother you if you sit with it. The flaw is fixed. Microsoft has said nothing about whether it was used against any tenant before the fix, and there is no breach notification or incident report attached. That is the honest limit of a transparency CVE. You are being told a serious flaw existed in a service your tenant depends on, and you cannot independently check whether your tenant was touched.

That is the real work this CVE leaves behind, and it is not patching. It is verification, of the limited kind that is actually available to you.

Four things worth a look

These are checks, not a runbook. Nothing here proves a clean bill of health. Each one is a place where an abused device registration would leave a trace.

Check What you are looking for
Devices registered recently In the Entra admin center, filter devices by registration date and look for entries you do not recognize, especially ones with no owner you can name.
Sign in logs for the Device Registration Service Unusual registration events, or registrations outside your normal onboarding pattern.
Device code flow in Conditional Access Microsoft's own Storm-2372 guidance is to block the device code flow for everyone except documented exceptions. If you never set that policy, this is a good week to write it.
Registered devices with no compliance state Devices that are registered but never became compliant or joined, sitting in the directory with trust and no owner.

If any of those turn up something odd, the next step is your normal incident process, not a patch. Revoke refresh tokens, remove the device, and work from your sign in logs. If they turn up nothing, that is what a clean check looks like. It is not proof nothing happened, and saying so is part of doing the job honestly.

The durable point

Cloud service CVEs invert the patching habit. There is no Patch Tuesday for a service you do not run. The vendor fixes it, tells you after, and the work that remains is deciding whether the flaw touched you, with less evidence than you would like.

That is worth getting used to, because this is the shape of more advisories to come. Microsoft publishes these under its cloud service CVE program precisely because the old model, where a CVE meant a patch you install, does not cover services. The score still matters. It tells you how seriously to take the component. But the action it demands is different: read it, check what you can check, and keep the record of what you checked.

Sources