A Sunday patch costs money. Guards, support engineers and deployment teams have to be mobilized outside working hours, and companies only accept it when waiting until Tuesday costs more than breaking the weekend. Microsoft did so yesterday with versions 8.0.33, 9.0.22, 10.0.14 and RC3 of 11, all of them for the .NET framework and all released out of band according to the release note, to patch a buffer overflow in the Brotli decoder built into the framework.
The flaw is tracked as CVE-2026-111497 and allows an unauthenticated attacker to read or corrupt the process memory, which can lead to data leakage, service outage or privilege escalation, as reported in the dotnet/runtime advisory. This is not an isolated development: it is the same defect that forced Node.js to release new versions on Wednesday the 21st, and the .NET Brotli case adds its own twist because Microsoft rates the impact more mildly than the rest of the ecosystem.
The entry point is not the framework in the abstract, but three very specific decompression routes: ASP.NET Core request decompression, HttpClient automatic response decompression and the Brotli APIs in System.IO.Compression. The first two are not enabled by default, which reduces the exposed universe without eliminating it. Anyone who has touched those options has work to do this Monday.
From an invalid block to an eight-byte buffer
The flaw did not originate in Redmond. It lives in Google Brotli, the library used by almost every web server to serve content with `Content-Encoding: br`, and it starts with a state-management error when reading the compressed stream.
A malformed stream can declare a block type change that its own structure invalidates. The decoder interpreted it as a request for more input and kept writing into an internal eight-byte buffer, according to the technical analysis of the flaw.
The affected versions range from 1.0.3 to 1.2.0, and the upstream fix arrived in 1.0.10, 1.1.1 and 1.2.1. The public CVE score is 9.2 out of 10 on CVSS v4, with a far from trivial attack requirement: the attacker must control the compressed data that reaches the decoder.
Microsoft rates the case as high and describes the impact as privilege escalation rather than remote code execution, a difference in criteria worth keeping in mind: the .NET Brotli flaw is prioritized for what it allows, not for the label it receives.
The comparison with last week is unavoidable. The critical PHP flaw went down in recent history as php2shell and this one repeats the pattern with another component: nobody installs Brotli by hand, because it arrives hidden inside the runtime you already use, the container you deploy or the package you declare as a dependency. That is why a library CVE ends up turning into four framework releases on any given Sunday.
What changes in each version?
The package is short on branches and long on consequences. Four product lines get a new runtime, and two of them are two weeks away from losing support, so the decision of what to install this Monday also conditions the migration that comes next.
| Line |
Runtime |
SDK that includes it |
What it adds |
| .NET 8 |
8.0.33 |
8.0.133 and 8.0.427 |
Brotli, tarballs and SocketAsyncEventArgs |
| .NET 9 |
9.0.22 |
9.0.123 and 9.0.320 |
The same fixes in its branch |
| .NET 10 |
10.0.14 |
10.0.114 and 10.0.403 |
Adds the Blazor debugger and BrowserDebugHost |
| .NET 11 |
RC3, 11.0.0-rc.3.26522.148 |
11.0.100-rc.3.26522.148 |
The four fixes in the November release candidate |
The second CVE on the list almost goes unnoticed and is the one that most troubles those who package. CVE-2026-114148 affects the ownership metadata of the tarballs for Linux, where misconfigured permissions let a local attacker modify files and gain privileges. Branches 8 and 9 inherit it in the ASP.NET Core distributions shipped as tarballs, even among those who do not compile the framework from source code. Downloading the new file does not by itself fix an installation that already exists.
The two warnings that return to the list
Reappearing is the awkward part of Sunday. Two advisories that were fixed on October 13 are back on the list because that release did not fully close either of them: CVE-2026-70197, memory corruption in SocketAsyncEventArgs, and CVE-2026-102870, remote execution through the Blazor WebAssembly debugger.
The release note for the update of .NET 10 admits that this new release also fixes the packaging of BrowserDebugHost, which was what was missing for the debugger patch to actually work on real installations.
For anyone who already applied Tuesday the 13th, the practical consequence is that they have to update again: the official recommendation is to install this release even if the monthly patch is in place. That detail also explains the insistence on restarting processes.
A new runtime on disk does not change the memory of a process that has been running for days with the old libraries loaded. On an ASP.NET Core server, that can mean recycling the application pool or the entire container during the early hours.
Why doesn't the installed runtime save everyone?
The rule that causes the most work is the one for self-contained applications. When publishing with `–self-contained`, the executable carries its own copy of the runtime and of Brotli; installing 8.0.33 on the machine does not replace anything inside that folder, so the binary remains exposed until someone recompiles it.
With NativeAOT the same thing happens in a worse form, because the decoder is linked into the executable and there isn't even a file you can replace, as detailed in the publishing guide. Containers fall into the same trap when the image pins a specific tag: you have to rebuild and redeploy.
And then there is PowerShell, which is what the headline is really about. Each major version is built on a different version of .NET, and self-contained installations copy the runtime into their own directory, so patching the flaw means upgrading the PowerShell version, according to the relationship between versions. Windows PowerShell 5.1 is spared because it runs on .NET Framework, the same one the notice leaves out.
You also have to add container images, artifacts published in internal repositories and loose binaries that load their own runtime inside. That list is almost never in a single place and has no clear owner, and that is where the .NET Brotli case stops being a download and becomes a deployment chore.
Fifteen days for branches 8 and 9
There is a clock running in parallel that does not depend on this CVE. Branch 8 and branch 9 lose support on November 10, 2026, so these 8.0.33 and 9.0.22 are among the last releases they will see, according to the support schedule.
For a company that has not yet migrated, Sunday's patch is a reasonable excuse not to invest in a branch that expires in two weeks and to jump to version 10, with support until November 2028. The jump is not free, because packages have to be reviewed and serialization tested.
Even so, sustaining two branches at once costs more than planning a migration with a date, and out-of-band patches exist precisely because the schedule does not allow delays. A branch in maintenance only receives security patches.
A Sunday that costs more than a Tuesday
The out-of-band release is measured by what it prevents, not by what it publishes. There are no public signs of real exploitation, and the status of the flaw in the Node ecosystem is that of a working proof of concept with no confirmed campaign to date. The calculation depends on each house: whoever decompresses third-party Brotli in an exposed service has a short window, and whoever only serves static files with another server can breathe.
The patching discipline is more of a routine than an alert, and it is worth treating it as a weekly review, just as current affairs are reviewed in the key stories of the week of October 23.
With two branches expiring in fifteen days, the pending work is no longer just installing four new numbers, but deciding which version the code will live in next year, and that decision is not solved by any Sunday patch.