Forgot your password?
typodupeerror
Bug Microsoft

Microsoft Breaks Another Patch Tuesday Record (bleepingcomputer.com) 63

Microsoft's September 2026 Patch Tuesday is its largest ever, fixing a record 966 vulnerabilities, including 105 rated critical and two zero-days already being exploited in attacks. BleepingComputer reports: This Patch Tuesday addresses 105 "Critical" vulnerabilities, 81 of which are remote code execution, 20 are elevation of privileges, 2 are information disclosure, and 1 security feature bypass.

The approximate number of bugs in each vulnerability category is listed below:
- 438 Elevation of Privilege Vulnerabilities
- 19 Security Feature Bypass Vulnerabilities
- 258 Remote Code Execution Vulnerabilities
- 173 Information Disclosure Vulnerabilities
- 56 Denial of Service Vulnerabilities
- 16 Spoofing Vulnerabilities
Last month, Microsoft's Patch Tuesday updates fixed 570 security flaws, following 400 vulnerabilities patched in August.

Microsoft Breaks Another Patch Tuesday Record

Comments Filter:
  • by gweihir ( 88907 ) on Wednesday September 09, 2026 @12:11PM (#66330770)

    That seems very bad even for Microsoft. Have they been vibe-coding or what?

    • That seems very bad even for Microsoft. Have they been vibe-coding or what?

      Yes, but only the patches.

      They found vibe-coding the core code only created..more patches.

      Which someone finally bitch-slapped the shit out of the marketeer who was selling that crap as a good thing.

      • by gweihir ( 88907 )

        Yes, probably. There is evidence making patches with LLMs often opens other issues, unless you review very carefully.

        • by gweihir ( 88907 )

          That said, with the mountain of technological debt that MS is lugging around, we may have a case here where patching by LLM (or even manually) cannot bring the number of bugs down or reduce their severity, but can only make things worse. This was, as far as I know, only a theoretical scenario for any mainstream software so far. Well, looks like MS may have done something for Science. If so, the only thing that can fix this is throwing it away and starting over. With a time-frame of 10-20 years or longer. Th

          • If so, the only thing that can fix this is throwing it away and starting over. With a time-frame of 10-20 years or longer. That will not go well.

            That go-back-and-rebuild-better will never happen for any currently deployed aspect of our global civilization again. At least not by human choice. It would only happen after some tangible event like some silicon-melting alien EMP attack or circuit-devouring nanobot swarm goes rogue.

            It's the same thing as that quote that goes something like, "The economy can stay irrational longer than you can stay solvent".
            The data centers can stay cycle through coprophagic tokens faster/longer than you can stay solvent.

            Tw

            • by jhoegl ( 638955 )
              Truth is, MS has been outsourcing their coding for over 20 years now.
              And the next truth is, they dont care about OS anymore. Cloud is all they care about. If you notice, not much has been innovated or iterated on Windows OS in the past 5-10 years. Its been more of a maintain and then slap in some online related bullshit to feed their ad revenue, or link to their online services that they want you to eventually move to.
              MS is not doing much as a company in the home space, except to try and pull as much info
              • by gweihir ( 88907 )

                But here is the thing: MS cannot run their cloud competently either, with several really bad hacks now and outages becoming more frequent. Without Windows, there is not a lot of reasons left to use the MS cloud. Exchange is one, but that is getting worse and worse as well.

                • by jhoegl ( 638955 )
                  Yes, cloud is where the continued slop is. They change names, move things around, and attempt to show they are "doing things", but really its just a bad UI, with terrible support, and shitty documentation while they continue to mess with fees and subscriptions to pull more money from companies and individuals who subscribe.
            • by shanen ( 462549 )

              Scarcely seems worth the effort in today's Slashdot, even when the FP is relatively good... But the aspect I've been failing to find so far in the discussion involves a different aspect or angle on that question. Something like "Are they sure?" or "How did they verify the bugs? Did they just ask another AI or were they able to construct test exploits?"

              The recursive race to the bottom of bug fixes causing as many or more bugs than they cure is important. Sure, but... I suppose the worst case would be if the

            • Slow and deliberate, see the Linux operating system, an actual operating system. Microsoft Windows stopped being an operating system a few decades ago. It's a collection of applications written by diverse teams who have no incentive to make one unified set of code with oversight in the major features. This coming from a guy who has been involved in software development since the 1980s. I've used a multitude of operating systems from real time through windows and of course Linux. For my embedded board space,
            • by gweihir ( 88907 )

              Apple has done it with OSX. Hence it can be done. It is hard though, slow and expensive.

              For MS to try this (if they do they very likely will fail, they have proven time and again they cannot even do a filesystem), this would require the cost of using Windows to become extreme. It is on its way there, but relatively slow. The alternate scenario is that anybody that needs reliability moves to Linux or one of the free BSDs and Windows just gets worse and worse.

              • Apple has done it with OSX.

                What? Apple did it with OSX, which is based on NeXTStep, which is based on Mach and FreeBSD? They absolutely did not do it with OSX. They dusted off NeXTStep, fucked up the interface, and made the necessary technical improvements to its ancient legacy codebase to make it function in the modern world... no better than any other legacy OS.

                • by gweihir ( 88907 )

                  It was not me that claimed a rebuild was needed. What is needed is a different base that is not fucked up. It can be made from existing stuff.

                  • What existing stuff is not fucked up? We need a software architecture that puts security first, whether that's microkernel-based or whatever else. And we need to rethink the way we handle speculative execution as well. We need substantive changes, not further rearrangement of deck chairs and a declaration of victory.

                    • by gweihir ( 88907 )

                      Well. At this time, the Linux kernel and the x-BSD kernels are not fucked up or not too much so and they are all actively and competently maintained. I do have some concerns about how the Linux kernel is growing, but that seems to not be too bad yet. The typical GNU user-space (without crap like systemd) is also in pretty reasonable shape, at least if you do competent system admistration.

                      I do agree that we cannot declare victory. We will have to be very careful and especially keep adversarial security resea

            • Rebuild is never as good as testing in a live environment. It makes you wonder those if Microsoft is patching 966 and Apple is patching 2 a month or so.. what is Apple and others missing. There has to be attack vectors targeted towards their systems as well.

          • by taustin ( 171655 )

            I recall seeing a study many years ago into how likely a bug fix was to introduce a bugs. Their conclusion was that the more lines of code you have, the more likely that was to happen. The numbers given were that at one millions lines of code, each bug fix would, on average, introduce 1.2 new bugs. Windows, at that point (XP?) had 400 million lines of code.

            • by gweihir ( 88907 )

              This has been a low-key research issue for as long as large software systems are around, hence for 30-40 years or so. LoC is not a good metric though. The problem we have if they are using LLMs for patch generation is that apparently LLM-written patches will introduce many more new bugs than human-made ones and apparently even regressions. We have no numbers on this, but some experiments had really bad effects even on smaller projects. Obviously, LLM code being hard to review for humans makes things worse.

              A

    • That seems very bad even for Microsoft. Have they been vibe-coding or what?

      More than likely these vulnerabilities have been there for a while, but they started to use AI to find them. Hence why they were able to find so many in this cycle.

      • by taustin ( 171655 )

        Even Linus has recognized the value of using AI to find vulnerabilities - provided that actual programmers review every single line before submitting it.

        • by gweihir ( 88907 )

          Well, you have to do it because the attackers have LLMs as well. Overall negative worth and the situation would be better without AI (as so many other things), but since AI is available you have to do this crap. What you should most definitely not do is let an LLM write the patch. That has a very high risk of causing more problem or being only a partial fix.

      • by antdude ( 79039 )

        Like in Linux kernels. So many!!!!!!

    • You joke of course, but finding new exploits with automation a real thing, scary effective with GPT-6 Astra as it recently scored a 100% on ExploitBench. Literally, the method of measuring how good it is has become insufficient. Anyway, we know Microsoft is automating these same kinds of tasks internally as well. Should we really complain that they aren't taking months or years? I think verifying AI patches actually work is the key point here, for both security and reliability, where the code came from does
    • by AmiMoJo ( 196126 )

      Nah, it's just the result of them using AI tools to find vulnerabilities. The thing to be worried about is that nobody is publicly throwing as many AI tokens at Linux and other critical open source code. I say publicly because I'm sure privately many security agencies and criminal enterprises (but I repeat myself) are doing it.

    • by Z80a ( 971949 )

      They're "vibe-fixing" it now.
      LLMs are pretty great at finding exploits, and by pretty great i mean they do more than 0% of success and you can engineer around it.
      It don't matter if the LLM is only correct 1% of the time, because you can test every exploit, and the ones that actually work, well, they go to the list to be checked by humans.
      So if it's the 1% i'm talking about and they "find" 10000 exploits, that's still 100 real exploits.
      It probably should be a mandatory build step when vibe coding.
      Like you le

  • Also took down the PACS server at our local hospital. The D drive didn't mount after this update.

    • Any luck with that? I see a bunch of changes that could touch it. AD, DNS, credentialguard, LSA, Netlogon...
    • I wonder what the record is for most bugs introduced on a patch Tuesday?

    • Hospital huh? I'd be impressed if PACS isn't running on a 15 year old Windows 10 Pro potato so the business suits can pad their paychecks for a new iPhone every year.
  • by smooth wombat ( 796938 ) on Wednesday September 09, 2026 @12:32PM (#66330812) Journal

    You could have stopped at Microsoft Breaks. We just found a previous update from Microsoft breaks the cellular connection in some of our HP laptops. The update flips the IMEI back to the default. We had to get a patch from HP to fix the update.

  • They are FEATURES designed to share all your data with the entire world. People just need to figure out how.

  • by whoever57 ( 658626 ) on Wednesday September 09, 2026 @01:56PM (#66330962) Journal

    I have a Windows 11 laptop that I use occasionally. My daily driver runs Linux.

    I'll keep the Windows 11 laptop turned off for a few days for any fallout to occur and bad patches to be pulled.

  • by ruddk ( 5153113 )

    I wonder how many things they break. :(

  • The fact that there were over 100 patches and 80 of them were for remotely exploitable bugs. The average ratio had been exactly the reverse the entire lifetime of the web for all OS's until now.

  • I'm loving Linux more and more. I have Linux Mint Cinnamon on 2 Laptops and Linux Zorin on another. Now Windows 11 is my bottom third OS only because there's a few Windws Apps still needed. It's so relaxing to NOT to have Windows stop me for very rude updates, in the middle of my work that can take hours to setup, reboot then reboot again.

    • I also use Mint and have noticed the updates there are running a higher rate than usual too. The AI can firehose end product around so effectively that everyone is finding vulnerabilities. Apple is not exempt either.

  • The new AI tools are making it easier to find problems

  • Update on one computer nuked internet access.

    On another, it nuked the camera, PIN, password, the password recovery questions. Wouldn't let me past the login screen.

    But it served up advertisements on the login screen. Took a week with my IT guy to fix it. Still not quite right, keeps nagging me about Teams and Onedrive.

  • I would not be proud to know that that many bugs exist in my code.
  • Our promise: you don't need a fan to cool your computer down - all those holes make passive cooling a reality!

We don't know one millionth of one percent about anything.

Working...