← Blog

November 2026 Windows Server updates: what to fix first, and what might break

This month, domain controllers go first, along with any server running DNS or DHCP. Microsoft’s November 2026 Patch Tuesday updates, released on 10 November, include four flaws in core infrastructure services that an attacker can use over the network with no sign-in at all, each rated 9.8 out of 10: in LDAP, Kerberos and Windows DNS, which ship in every version of Windows, and in DHCP Server.

Microsoft’s release fixes more than 1,000 vulnerabilities across its products. On 10 November, Microsoft identified one of them as already exploited: a flaw in Outlook, fixed by the Office update, not the Windows one. This post covers what matters for Windows servers: what to patch first, what the update changes, what it repairs from October, including the Exchange problem, and a certificate change from 16 November.

What to prioritise

Domain controllers, DNS and DHCP servers first. Four of the month’s most serious flaws need no sign-in and no action from anyone: an attacker only has to reach the vulnerable service over the network. Three of them, in LDAP, DNS and Kerberos, affect every supported version of Windows, so every server needs this update. Domain controllers go first because they provide LDAP and Kerberos, and often DNS. Servers running DNS or DHCP belong in the same first round, wherever those roles are installed, and so do Exchange servers that receive mail from the internet (see below). Microsoft doesn’t say whether the DNS flaw is reached through the DNS Server role, so don’t treat a server as safe because it isn’t a DNS server.

CVEComponentCVSSSign-in needed?What it allows
CVE-2026-119958LDAP9.8NoRemote code execution, through crafted LDAP calls
CVE-2026-88218DNS9.8NoRemote code execution
CVE-2026-88243DHCP Server9.8NoRemote code execution
CVE-2026-85844Kerberos9.8NoSYSTEM privileges on the domain controller
CVE-2026-118146Active Directory Domain Services8.8Yes, a low-privileged accountRemote code execution

Microsoft had not reported exploitation of any of these five as of 10 November. Microsoft rates all four 9.8s “exploitation less likely”, and the Active Directory one “more likely”. Keep domain controllers, DHCP servers and internal DNS servers off the public internet. If you run DNS servers that answer the internet on purpose, for your own domains, patch them promptly and allow recursive lookups only for your own clients. Treat the order in this post as a starting point: what goes first depends on which services you run and who can reach them.

Services that handle TLS themselves. CVE-2026-119952 is a remote code execution flaw in Schannel, the part of Windows that handles TLS. An attacker needs no sign-in, only the ability to send crafted packets to a service that uses it. It depends on specific configurations, which is why it scores 8.1 and not higher, but Microsoft rates exploitation “more likely”. Servers whose IIS sites, Remote Desktop or other services accept TLS connections from outside come next. If HTTPS ends on a reverse proxy or load balancer that doesn’t run Windows, the server behind it is less exposed to this one.

Others that need no sign-in. CVE-2026-119961 is a remote code execution flaw in IIS, rated 8.1, which gives web servers a second reason to go early. CVE-2026-119962 is in CredSSP, the protocol Remote Desktop uses for Network Level Authentication: also 8.1 and no sign-in needed, which matters for servers reachable over Remote Desktop. CVE-2026-113862, in Windows NAT, is 8.1 and rated “exploitation more likely”. CVE-2026-88234, in Windows iSCSI, is rated 9.8 and allows remote code execution without a sign-in; it matters most on servers involved in iSCSI storage. And if the server runs Windows Deployment Services (CVE-2026-119951) or has Message Queuing installed (CVE-2026-119963 and four others), it has a no-sign-in flaw of its own this month, each rated 8.1.

Hyper-V hosts. If you run Hyper-V hosts with virtual machines for different users or customers, patch the hosts early too. Two November flaws, CVE-2026-113861 and CVE-2026-119947, can let code running inside a guest virtual machine execute code on the host, with the privileges of the VM worker process. Updating Windows inside the guests doesn’t patch the host.

On-premises Exchange, for two reasons. CVE-2026-116928 is a flaw in Windows Search, rated 9.8. On an Exchange server, an email with a crafted attachment can trigger it when Exchange indexes the message, without anyone opening it, which is why Exchange servers that receive internet mail belong in the first round. The fix is in the Windows update, so it’s the Exchange server’s operating system that needs patching. Separately, the November Exchange security update, SU12 for Exchange Server Subscription Edition (KB5133453), fixes nine Exchange flaws, including CVE-2026-118207, a remote code execution flaw: any mailbox account is enough to exploit it, and Microsoft rates exploitation “more likely”. Microsoft says to install the update on every Exchange server, including hybrid servers used only for management, and on any server or workstation with the Exchange Management Tools installed. After installing it on an Exchange server, restart it, check that all Exchange services have started and run Microsoft’s Exchange Health Checker. On servers that handle mail, also test Outlook on the web sign-in and mail flow. Exchange 2016 and 2019 no longer receive fixes at all, and Microsoft now describes them as permanently vulnerable. If you still run either, start the migration to Subscription Edition now.

The exploited one isn’t in Windows. CVE-2026-124194 is an Outlook flaw: when someone opens a specially crafted message file, Outlook connects to an attacker’s server without its usual warning, which can expose the user’s NTLM authentication data. Previewing the message isn’t enough; it has to be opened. It’s rated only “moderate”, but on 10 November Microsoft identified it as already exploited, and CISA has added it to its Known Exploited Vulnerabilities catalogue. It’s fixed by the Office update, not the Windows one. Office doesn’t belong on a server, so strictly speaking this isn’t a server problem. In practice, though, plenty of people rent a Windows VPS from a cloud provider and use it as their everyday desktop, with Office installed, because most providers offer Windows Server rather than Windows 11. We include it for them: it’s the one flaw this month that attackers are already using. If that’s how you use a server, install the Office update now.

Other Microsoft software on the server. The Windows update doesn’t cover SQL Server or .NET. If the server runs SQL Server, check for a November security update for its version and servicing branch. If it runs .NET applications, install the November .NET updates too, and see Deadlines for the versions that have just reached end of support.

Changes to check

Signed PowerShell scripts. Windows now checks a script’s text encoding more strictly when it signs a script or verifies a signature. A signed script whose text Windows can’t read cleanly, because it contains invalid UTF-8 or was saved in a different code page, can stop validating after the update. The signs, according to KB5135999: Get-AuthenticodeSignature reports NotSigned for a script that was signed, or, where App Control for Business is in use, the script runs in ConstrainedLanguage mode. Those signs can have other causes, so before rolling the update out, run the scanning script from Microsoft’s article over the folders that hold signed scripts used by scheduled jobs. Run it in Windows PowerShell 5.1 with -Mode Scan, which only reads files. For your own scripts, convert a copy to UTF-8 with a byte order mark from its original encoding, check the text, re-sign it and confirm the signature shows as Valid. For scripts signed by a supplier, ask the supplier for a corrected copy. In the meantime, the script’s -Mode Repair can produce a copy that keeps the supplier’s signature; use it only if the copy shows as Valid. After updating a test server, run the jobs under their usual account and policy. This applies to Server 2016, 2019, 2022 and 2025.

NTLM message signing, phase one. To address CVE-2026-118737, a remote code execution flaw in the Windows RPC runtime (rated 8.1, no sign-in needed, but hard to exploit), Windows now offers NTLM message signing by default on every connection except anonymous and guest ones. This also reduces exposure to NTLM relay attacks. Nothing is refused yet: if the other end doesn’t sign, the connection carries on as before. The second phase is the one to plan for. From May 2027, Windows servers will require signing by default and reject NTLM clients that don’t sign. That leaves about six months to find them. KB5135557 has the details:

  • On Server 2019 and later, enhanced NTLM logging records connections in the Microsoft-Windows-NTLM/Operational log, and Microsoft says it’s on by default. On the receiving server, review events 4022 and 4023. These events need Server 2019 or later, and on Server 2019 and 2022 the logging only started with the October 2026 update, so the history is short. An empty log on Server 2016 proves nothing.
  • Event severity alone doesn’t tell you whether signing was negotiated. Use the script in KB5135557 to find connections missing the SIGN flag. When checking a single event against Microsoft’s example of a ready connection, look for SIGN (0x10), KEY_EXCH (0x40000000) and a session key status of Present.
  • Once auditing over a representative period, including scheduled jobs, shows no unsigned connections, you can require signing early by setting ServerRequireNtlmSign to 1 under HKLM\SYSTEM\CurrentControlSet\Control\Lsa\MSV1_0. No restart is needed.
  • Kerberos isn’t affected. Old devices and applications that still use NTLM are where to look.

This is separate from the NTLMv1 change we mentioned in October’s post.

Smaller changes.

  • Windows Deployment Services. Import-WdsInstallImage now needs administrator rights and fails with ACCESS_DENIED without them.
  • Installation media. If you build or update Windows Server installation media with the November update, include the boot.stl file that matches the image’s Windows version and architecture. Without it, servers can fail to start from that media with error 0xc0430001. This only concerns people preparing installation media; installing the update on a running server needs nothing extra.

A known issue from October: SQL Server linked servers. On 11 November, Microsoft confirmed that since the October 2026 update, connections created by SQL Server linked server queries that use OLE DB providers might stay open after the query completes. Connections to external databases such as Oracle and PostgreSQL build up, and the server can slow down over time. The November update doesn’t fix it. Microsoft lists Server 2016, 2019, 2022 and 2025 as affected, and says an out-of-band update will follow in the next few days. Until then, Microsoft Support for business offers a mitigation. Restarting the SQL Server service, or closing idle connections, gives temporary relief, but it can interrupt running queries and the problem comes back. If you use linked servers to databases other than SQL Server, watch the number of connections to them.

Activation commands (fixed on 13 November). After the November update, some activation and licensing commands run with slmgr could fail with error 0x80041008 (“Invalid parameter (SWbemObjectEx)”) on Server 2025, 2022 and 2019, for example when installing a product key (slmgr /ipk) or pointing a server at a KMS host (slmgr /skms-domain). Servers that were already activated stayed activated. On 13 November, Microsoft fixed it with out-of-band updates that are only available from the Microsoft Update Catalog: KB5136489 for Server 2025 (build 26100.33828), KB5136487 for Server 2022 (20348.5873) and KB5136490 for Server 2019 (17763.9488). They’re cumulative and include the November security update. You only need them where you run these commands, such as servers you build images from.

Printing documents with zero-size images. Microsoft has confirmed in the Microsoft 365 admin centre, as reported by Born’s IT and Windows blog, that after the November update printing can fail on all four Server versions when a document contains an image with a height or width of zero at the selected resolution. It was seen with Microsoft Word, and other applications may be affected; print jobs can disappear without printing. There’s no fix or workaround yet. If a server acts as a print server, or hosts Remote Desktop users who print, test printing after updating.

An existing limitation: WSUS. Synchronisation error details are still missing from the WSUS console, and Microsoft still lists this as a known issue on Server 2025 and 2022. Microsoft removed them a year ago to address CVE-2025-59287 and hasn’t brought them back, so when a synchronisation fails, the console no longer tells you why. Look in SoftwareDistribution.log on the WSUS server instead, and after each sync check that the new month’s updates actually appear in WSUS.

As of 17 November 2026, Microsoft has confirmed two problems caused by the November updates: the activation one, now fixed, and printing, which isn’t yet. The SQL Server issue dates from October, and its fix hasn’t been released yet. If a new problem appears, our guide to rolling back a Windows Server update covers the options, starting with the least drastic.

Fixed from October

Exchange and the RPC Endpoint Mapper. The problem that stopped people signing in to Outlook on the web and left mail stuck in queues after the October updates is resolved. Microsoft marked it “Resolved” on its release health pages on 10 November, by KB5131080 on Server 2025 and KB5131090 on Server 2022, with the same fix in the Server 2019 and 2016 updates. It’s the Windows update that fixes it, not the Exchange one, and the two can be installed in either order.

If you applied the EpmapSavedContextCap registry workaround, Microsoft says it’s no longer needed once the November update is installed. You can leave it in place or remove it.

Other repairs.

  • Adding a printer published in Active Directory through Settings failed with error 87 after the October update. Fixed on all four versions.
  • Remote Desktop shadow connections on Server 2016 failed with “Unspecified error” after the October update. Fixed.
  • RD Web Access pages that failed to load in Edge or Chrome where browser policy disables XSLT now load, because the pages are built on the server. Check any customised RD Web pages after updating.
  • In-place upgrades from Server 2016 to Server 2025 using October installation media could fail the compatibility check with error 0x80070003 - 0x40008. Fixed in the Server 2025 update. If you hit this while upgrading a server that can be upgraded in place, following our Windows Server 2016 end of life guide, use media that includes the November update. Exchange servers can’t be upgraded this way: Microsoft doesn’t support an in-place Windows upgrade with Exchange installed.
  • USB Audio Class 1.0 devices that showed “Code 10” or produced no sound since September: the remaining cases are fixed on Server 2025, 2019 and 2016. Server 2022 had its fix in October.
  • Cuba’s clocks. Cuba stays on UTC-04:00 through the winter, with no change in November 2026 or March 2027. The update corrects the time zone rules.
  • Root certificates. The out-of-band updates of 26 October (KB5134137 for Server 2025, KB5134140 for 2022, KB5134124 for 2019, KB5134123 for 2016) removed seven eMudhra-related certificates from the Windows trust list and delivered an updated list of trusted root certificates to servers that can’t fetch it from Microsoft themselves. The November updates include both. More removals start on 16 November; see Deadlines.

Deadlines

eMudhra root certificates leave Windows from 16 November. Microsoft is withdrawing trust in eMudhra, a certificate authority, after eight of its intermediate certificate authority keys were compromised. The November update already removes seven eMudhra-related certificates. In an update to its announcement (message centre MC1529467), Microsoft says that from 16 November it is removing the remaining ten eMudhra roots, including emSign Root CA - G1, through Windows’ automatic trust-list updates, so a server can be fully patched and still see this change. Check whether any website, application or service you run uses a certificate that chains to eMudhra or to a root named emSign. Replace the certificate, or make sure it chains to another trusted root, as soon as possible. Servers that don’t receive automatic trust-list updates, such as those without internet access, won’t get the change on their own: follow Microsoft’s guidance for distributing updated trust lists to them.

Windows Server 2016 has two scheduled security updates left: 8 December 2026 and 12 January 2027. After that, Standard and Datacenter editions can get Extended Security Updates, at a cost, for up to three more years. Essentials has no Extended Security Updates option, so it needs replacing before 12 January. If you still run Server 2016, our Windows Server 2016 end of life guide covers the options.

Windows Server 2022 is now in extended support. November is its first monthly update since mainstream support ended in October. Security updates continue, at no extra cost, until 14 October 2031.

.NET 8, .NET 9, PowerShell 7.4 and PowerShell 7.5 reached end of support on 10 November 2026. If an application on your server runs on .NET 8 or 9, or your scripts run on PowerShell 7.4 or 7.5, they’ve had their last security fixes. Move them to .NET 10 and PowerShell 7.6, both long-term releases supported until November 2028. For an application you didn’t write, ask the supplier for a version that runs on .NET 10; for one you maintain, retarget, test and redeploy it. Installing .NET 10 on the server on its own doesn’t move an application across.

NTLM signing becomes required by default on servers in May 2027, as described above.

A short checklist

  • Confirm you have a recent backup you can restore before patching. For domain controllers, the backup must include system state; for Exchange, use Exchange-aware backup software. A virtual machine snapshot isn’t a substitute for either.
  • Patch domain controllers, DNS and DHCP servers, and Exchange servers that receive internet mail first, then servers that accept TLS or web traffic from the internet and Hyper-V hosts, then the rest. Adjust the order to what you run and who can reach it.
  • Where you have more than one domain controller, update them one at a time, and check DNS, authentication and replication before moving to the next.
  • If you run Exchange, install the Windows update on the Exchange servers: it fixes both the Windows Search flaw and October’s Exchange problem. On Exchange Server Subscription Edition, install SU12 (KB5133453) too, including on management-only servers and on workstations with the Management Tools, then restart, check the Exchange services and run the Exchange Health Checker. Exchange 2016 and 2019 get no Exchange update.
  • Confirm the update installed by checking the build number in the table below.
  • Check for eMudhra certificates in anything you run, as soon as possible.
  • Scan signed PowerShell scripts before rolling the update out, and test scheduled jobs afterwards.
  • If SQL Server uses linked servers to Oracle, PostgreSQL or other non-SQL Server databases, watch the connections to them until Microsoft’s out-of-band fix arrives.
  • If you build server images or change product keys or KMS settings, install Microsoft’s out-of-band fix for your version first (KB5136489, KB5136487 or KB5136490).
  • Test printing on print servers and Remote Desktop hosts after updating.
  • Review NTLM events 4022 and 4023 on Server 2019 and later for connections without signing, ahead of May 2027.
  • Check Microsoft’s Windows release health page during the week after an update for newly acknowledged issues.

November builds by Windows Server version

VersionUpdateBuild after the November update
Windows Server 2025KB513108026100.33822
Windows Server 2022KB513109020348.5870
Windows Server 2019KB513108517763.9487
Windows Server 2016KB513129114393.9688

Servers with the activation fix installed show 26100.33828, 20348.5873 or 17763.9488 instead.

Before installing:

  • Server 2016: install the servicing stack update KB5131083 first. Microsoft warns the main update might otherwise not be offered or might fail to install. Windows Update handles this automatically; in WSUS, approve both KB5131083 and KB5131291.
  • Servers not updated since before July 2025: on Server 2016, 2019 or 2022, check the installation prerequisites in the KB article first.

Two exceptions for servers on hotpatching:

  • Server 2025 machines enrolled in hotpatching get KB5131077 instead, which shows build 26100.33730. If they installed the out-of-band update of 26 October (KB5134137), though, they get the full November update, and hotpatching resumes after they install the January 2027 baseline. Microsoft’s notes for KB5131077 don’t list the Exchange fix, so on a hotpatched Exchange server, keep the EpmapSavedContextCap value until the full update is installed or Microsoft confirms the hotpatch carries the fix.
  • Server 2022 Datacenter: Azure Edition machines on hotpatching get KB5131090 this month as a baseline, and need a restart. Microsoft’s hotpatch calendar shows December 2026 as a restart baseline too.

Sources

This post is general information, not professional advice for your specific setup.

Rather not do this yourself?

Sorting the urgent from the noise every month, and patching in the right order, is part of what our Server Management service takes off your hands.

See how it works →