Microsoft October 2026 Patch Tuesday: a record 1,400+ CVEs, one exploited Windows Shell zero-day, and what to patch first

Microsoft's October 2026 Patch Tuesday, released on October 13, fixed more than 1,400 vulnerabilities: 1,410 by the count in Microsoft's own Security Update Guide. It's the largest monthly release Microsoft has shipped, beating September's record, and the first to pass 1,000 CVEs. Only one of the bugs is known to be under attack: a Windows Shell flaw that lets a malicious file slip past Mark of the Web warnings, which CISA added to its Known Exploited Vulnerabilities catalog the same day. For security teams, the volume itself is the story: Microsoft has tied the growth in volume to AI-assisted vulnerability research, and one of the exploited bug's credited finders used Anthropic's Claude. That makes this month a good snapshot of what frontier models are doing to the patch cycle.

This story is based on Microsoft's Security Update Guide and MSRC blog posts, and on analyses from the Zero Day Initiative, Tenable, BleepingComputer and Dark Reading, checked October 14, 2026.

The short version

• Size: 1,410 CVEs released on October 13 in Microsoft's Security Update Guide, excluding Azure Linux packages. Security firms count 1,407 to 1,415.
• Severity: 79 rated Critical in the guide; one of those is a republished Cisco Talos bug, leaving 78 that Microsoft issued.
• Exploited: one, CVE-2026-96951 (Windows Shell security feature bypass, CVSS 8.8). CISA added it to the Known Exploited Vulnerabilities (KEV) catalog on October 13, 2026.
• Publicly disclosed, not exploited: CVE-2026-108398 (.NET Framework, bypasses Windows Defender Application Control) and CVE-2026-92707 (MessagePack-CSharp denial of service, rated Moderate).
• Patch first: the Windows Shell fix, then internet-facing Exchange, SharePoint, SQL Server and Outlook systems.

How many vulnerabilities did Microsoft fix in October 2026?

It depends on who's counting. Microsoft's October 13 risk cohorts post gives no total. Its Security Update Guide returned 1,410 CVEs with an October 13 release date when queried on October 14, not counting Azure Linux (Mariner) open-source package entries. Eight more were published earlier in October for cloud services and Exchange Server, for 1,418 in the 2026-Oct release.

Last verified: October 14, 2026

Some outlets reported 1,118. That figure is the sum of the 15 risk cohorts, which don't cover every CVE. The trend is not in dispute: Infosecurity Magazine reported a record 974 CVEs in September's release, and Krebs on Security reported 570 in July. Tenable says elevation of privilege makes up 47.9% of October's bugs and remote code execution 22.3%.

What is the exploited Windows Shell zero-day?

CVE-2026-96951 is the only October bug Microsoft marks as exploited. Microsoft describes it as command injection in Windows Shell that lets an attacker bypass a security feature. Tenable says an attacker has to persuade a target to open a malicious file, which then skips Mark of the Web, the tag that triggers SmartScreen and Office Protected View. It's rated Important, not Critical. Tenable says it's the third exploited Windows Shell flaw of 2026.

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-96951 to its Known Exploited Vulnerabilities catalog on October 13, 2026, listing it as a "Microsoft Windows Command Injection Vulnerability" with a remediation due date of October 16, 2026. CISA's entry marks known ransomware use as "Unknown."

Microsoft's acknowledgments credit four finders: Riku Bamba and Yuya Chudo of Fujitsu, "yhw," and "Benoit Sevens using Claude with Anthropic." Dark Reading wrote that Claude "uncovered" the bug. Microsoft's credit is shared, and it doesn't describe how Claude was used. Microsoft hasn't said who exploited the bug.

The two publicly disclosed bugs aren't known to be exploited. Google Project Zero's James Forshaw is credited with the .NET Framework flaw. Some coverage called SharePoint bug CVE-2026-102866 a zero-day, but the Security Update Guide lists it as neither exploited nor publicly disclosed.

Does Microsoft say AI is behind the surge?

Yes, as a trend, though its October post doesn't credit AI for this month's count. In a May 12 MSRC note, Microsoft said advanced AI models "are part of the discovery picture and help to accelerate it" and that it expects releases to keep growing "for some time." In a July 9 blog post, Microsoft executive vice president Pavan Davuluri wrote that users would see "a higher volume of security updates," Krebs on Security reported.

On October 7, Microsoft vice president Taesoo Kim wrote that Microsoft's AI research team was behind 52 CVEs in the September release, using a multi-model scanning harness Microsoft calls MDASH. That's a small share of the total, and Microsoft hasn't said how many October CVEs came from AI tools. The October 13 post says only: "As AI-assisted research contributes to vulnerability discovery, this kind of approach may become more useful."

Others draw the same link with a caveat. "AI is certainly helping in bug discovery," ZDI's Dustin Childs wrote, "but we haven't seen the same increase in the exploitation of 0-days – yet." Fortra's Tyler Reguly told Dark Reading that under 1% of Microsoft's 2026 CVEs are known to be exploited. Anthropic's own scanning work is described in our Cyber Mission story.

What are Microsoft's risk cohorts?

They're 15 provisional groups of CVEs that call for the same question about your environment, such as "Windows filesystems, storage, and backup" (136 CVEs) or "Remote access and RDP" (65). You answer the question once per cohort instead of once per CVE. Microsoft says they are "not official Microsoft vulnerability categories," don't change update guidance, and "do not replace careful review of high-severity or actively exploited vulnerabilities."

What should admins patch first?

Childs' rule: "Patch what's proven, prioritized, and public, and treat everything else as this month's new normal." Drawing on Microsoft's ratings and the analysts above:

1. CVE-2026-96951 (Windows Shell). It's exploited and relies on someone opening a file, so it can arrive by email or chat.
2. Exchange Server. Tenable flags CVE-2026-70571 (Critical remote code execution, "Exploitation More Likely") and CVE-2026-70279 (spoofing, CVSS 9.3).
3. Outlook CVE-2026-70242 (CVSS 9.8, Critical). Microsoft says "simply viewing the message in the Outlook Reading Pane could trigger the vulnerability," though it rates exploitation less likely.
4. Internet-facing servers: Active Directory CVE-2026-70229, SharePoint CVE-2026-102866 (unauthenticated), SQL Server CVE-2026-65674, and Microsoft QUIC CVE-2026-105016 (CVSS 9.8).
5. The .NET Framework bypass (CVE-2026-108398) on machines that rely on Application Control.

Tenable also notes support endings on October 13: Windows 11 Home and Pro 24H2 and Office 2021 reached end of support, and Windows Server 2012 and 2012 R2 Extended Security Updates ended. Patch volume is also why "update everything" agents and exposed self-hosted AI servers need care; see our prompt injection explainer and our story on the PoeLLM botnet.

What's confirmed, reported and uncertain

• Confirmed (Microsoft): release date; Security Update Guide counts, severities and exploited/disclosed flags; CVE-2026-96951's finder credits; risk cohorts; Microsoft's AI statements from May, October 7 and October 13.
• Confirmed (CISA): CVE-2026-96951's addition to the KEV catalog on October 13, 2026, with an October 16 due date.
• Reported: total counts from ZDI, Tenable, BleepingComputer and Dark Reading; September's 974; Davuluri's July statement (via Krebs on Security); Dark Reading's description of Claude's role.
• Uncertain: how many October CVEs came from AI tools, how Claude was used on CVE-2026-96951, and who is exploiting it.