Between 1 March and 10 November 2026, the US Cybersecurity and Infrastructure Security Agency (CISA) added 250 vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog. Each one is a flaw with reliable evidence of exploitation in real attacks, not a theoretical weakness. We downloaded the catalog, counted every entry added in that period and classified each product by what it does. This page is the result: the monthly trend, the vendors, the product types attackers went after, and what a European company with 50 to 2,000 employees should change in its patch policy because of it. It is also the hub for our 2026 pieces on individual campaigns, from firewalls and VPNs in spring to the email gateways in autumn.
Key Takeaways
- 250 KEV entries in just over eight months. CISA added 295 entries in 2026 up to 10 November, already more than the 245 it added in the whole of 2025.
- September was the peak, with 43 entries, 18 of them firewalls, VPNs, gateways and other network edge devices.
- One entry in four (62 of 250) is a network edge or security appliance. Add remote access and IT management tools and the share rises to 79 entries, or 31.6%.
- Microsoft (30) and Cisco (18) lead the vendor list, but the catalog names 98 different vendors in the period. Patching only the big names leaves most of the list uncovered.
- Since June, most new entries come with a three-day deadline and a forensic triage requirement for US federal agencies. The message for everyone else: patch fast, then check whether you were already compromised.
Method: What We Counted and What KEV Does Not Tell You
The source is the KEV catalog in JSON format, published by CISA at cisa.gov. We used catalog version 2026.11.10, released on 10 November 2026, which contained 1,779 entries in total. We counted the entries whose dateAdded falls between 1 March and 10 November 2026 inclusive, and grouped them by month, by the vendor name CISA records and by product type.
The product types are our own classification, made by reading each product name and description: network edge and security appliances (firewalls, VPN and SSL VPN gateways, application delivery controllers, SD-WAN, email security gateways, routers and switches, and the consoles that manage them), remote access and IT management tools, operating systems and endpoint software, on-premises business and collaboration servers, web applications and the web stack, AI and automation platforms, developer tools and the software supply chain, virtualisation and backup, and OT, IoT and embedded products. Some products could sit in two groups; we placed each in the one that best describes where it runs and who can reach it.
Four limits are worth keeping in mind before reading the numbers:
- KEV has a US federal focus. CISA lists what matters to US federal agencies. Products common in European companies but rare in US government can be under-represented.
- It only lists confirmed exploitation. A vulnerability that is being exploited but has not been confirmed publicly is not in the catalog yet. Absence from KEV is not proof of safety.
- The date added is not the date of first exploitation. Some flaws were exploited for weeks before being listed; some are years old and were added only in 2026.
- The ransomware flag changes over time. CISA marks an entry as "Known" ransomware use when it has evidence, and it can update that field later. Recent entries are likely under-counted.
Anyone can repeat the count: download the JSON file, filter on dateAdded and group by vendorProject. If you get a different number, check the catalog version first: CISA updates the file several times a week.
The Monthly Trend: A Busy Autumn
| Month (2026) | KEV entries added | Of which network edge devices | Flagged "Known" ransomware use |
|---|---|---|---|
| March | 26 | 3 | 2 |
| April | 31 | 4 | 9 |
| May | 21 | 3 | 3 |
| June | 23 | 8 | 3 |
| July | 26 | 10 | 4 |
| August | 31 | 3 | 2 |
| September | 43 | 18 | 0 |
| October | 36 | 10 | 0 |
| November (1 to 10) | 13 | 3 | 0 |
| Total | 250 | 62 | 23 |
Source: CISA KEV catalog, version 2026.11.10. Product classification and counts by Orizon.
The average is about 28 new entries a month, but the distribution is uneven. From March to August the catalog grew by 21 to 31 entries a month. In September it jumped to 43, the highest month of the period, and October stayed high at 36. The first ten days of November already brought 13, including two Barracuda flaws and an OpenSSH vulnerability.
The year-on-year comparison is the more telling one. In the same window (1 March to 10 November) CISA added 176 entries in 2025, 124 in 2024 and 142 in 2023. The 2026 figure of 250 is 42% higher than in 2025. Counting from 1 January, 2026 stands at 295 entries, already above the 245 added in the whole of 2025. More confirmed exploitation, listed faster, means less time between "a patch exists" and "someone is using this against companies like yours".
Which Vendors Appear Most
| Vendor (as recorded by CISA) | Entries, March to November 2026 | Main products involved |
|---|---|---|
| Microsoft | 30 | Windows components (9), SharePoint (8), Defender (3), Exchange (2), Office (2), Internet Explorer (2), others |
| Cisco | 18 | Catalyst SD-WAN (7), Secure Firewall FMC, ASA and FTD (5), Secure Email Gateway (2), others |
| 9 | Chromium and its components (7), Pixel, Brotli | |
| Apple | 9 | iOS, iPadOS, macOS and other Apple platforms |
| Linux | 8 | Linux kernel |
| Fortinet | 8 | FortiClient EMS (2), FortiSandbox (2), FortiOS and related products (3), FortiMail (1) |
| Ivanti | 7 | Connect Secure (3), Endpoint Manager Mobile (2), Endpoint Manager, Sentry |
| Adobe | 7 | Acrobat and Reader (3), Commerce and Magento (3), ColdFusion |
| Citrix | 6 | NetScaler ADC and Gateway |
| SonicWall | 5 | SMA1000 appliances (4), SonicOS SSL VPN (1) |
Source: CISA KEV catalog, version 2026.11.10. Vendor names as recorded in the catalog; product groupings by Orizon.
The top ten vendors account for 107 entries, or 43% of the total. The other 143 entries are spread across 88 more vendor names, many of them with a single entry: CMS plugins, file transfer servers, help desk software, media servers, AI frameworks. That long tail is the practical problem. Most companies have a patch process for Windows, Office and the browser. Far fewer have one for the Joomla extension on the marketing site, the print management server or the open-source AI tool a team installed last spring.
Two products stand out on their own. SharePoint Server appears eight times in the period, which is why we gave it a separate article. Cisco Catalyst SD-WAN appears seven times, most of them in the Manager component that administrators reach through a web interface, a pattern we covered in our piece on management interfaces exposed to the internet.
What Kind of Product Gets Exploited
| Product type (Orizon classification) | Entries | Share | "Known" ransomware use |
|---|---|---|---|
| Network edge and security appliances | 62 | 24.8% | 6 |
| Operating systems, browsers and endpoint software | 52 | 20.8% | 2 |
| On-premises business and collaboration servers | 44 | 17.6% | 4 |
| Web applications, CMS and web stack | 33 | 13.2% | 1 |
| Remote access and IT management tools | 17 | 6.8% | 4 |
| AI, machine learning and automation platforms | 14 | 5.6% | 0 |
| Developer tools and software supply chain | 14 | 5.6% | 4 |
| OT, IoT, embedded and engineering software | 8 | 3.2% | 1 |
| Virtualisation and backup | 6 | 2.4% | 1 |
| Total | 250 | 100% | 23 |
Source: CISA KEV catalog, version 2026.11.10. Classification by Orizon, one category per entry.
Edge devices: one entry in four
Firewalls, VPN gateways, application delivery controllers, SD-WAN controllers, email gateways, routers and the consoles that manage them make up the largest group: 62 entries from 16 vendors. Cisco leads with 16, followed by Citrix and Fortinet (6 each), SonicWall (5), Check Point, Arista and Ivanti (4 each), Palo Alto Networks, Ubiquiti and MikroTik (3 each), F5 and Barracuda (2 each), and one each for D-Link, DD-WRT, Progress and Zyxel.
These devices share three traits that attackers value. They face the internet by design. They sit in front of everything else, often with credentials for the directory or the VPN users. And they are hard to monitor: most companies cannot install an endpoint agent on a firewall, so a compromise can go unnoticed for weeks. September alone brought 18 edge entries, including three NetScaler flaws, three in MikroTik RouterOS, two SonicWall SMA1000 flaws and a Cisco Secure Email Gateway flaw that can be triggered simply by sending an email.
If you add the 17 entries for remote access and IT management tools (SimpleHelp, ScreenConnect, N-able N-central, BeyondTrust Remote Support, Ivanti and Fortinet endpoint management, among others), 79 of the 250 entries, 31.6%, are products whose job is to give someone remote control of the network. That is the attacker's shortcut: one exploited box and they are inside, with the tools the IT team uses.
Servers you host yourself
The 44 entries for on-premises business and collaboration servers include SharePoint (8), Zimbra (3), PaperCut (3), Exchange (2), Oracle E-Business Suite and PeopleSoft, Atlassian, ONLYOFFICE, Collabora, ownCloud and IBM Aspera Faspex. The common thread is a server that a company runs itself and often publishes to the internet so that staff or partners can reach it. Moving these services to a managed cloud version does not remove all risk, but it removes the patching delay, which is where most of these entries hurt.
Web applications and the web stack
The 33 web entries are a reminder that the public website is part of the attack surface. WordPress core appears four times, Joomla extensions five times (JCE, SP Page Builder, Joomlack Page Builder, Balbooa Forms, iCagenda), Adobe Commerce and Magento three times, and cPanel or its plugins three times. These flaws fuelled the summer defacement wave against Italian sites that we described in a dedicated article.
The Ransomware Subset
23 of the 250 entries (9.2%) are flagged by CISA as "Known" to be used in ransomware campaigns. They include remote support and RMM tools (SimpleHelp twice, ConnectWise ScreenConnect, SolarWinds Web Help Desk), VPN and firewall products (Palo Alto Networks PAN-OS, Check Point Security Gateway, two SonicWall SMA1000 flaws, two in Cisco Secure Firewall Management Center), Exchange and SharePoint, VMware vCenter, JetBrains TeamCity (twice), PaperCut, Oracle PeopleSoft and PTC Windchill, the product lifecycle management platform used across manufacturing.
The pattern matches what incident responders describe: ransomware operators prefer a way in that gives them administrative reach quickly, such as a VPN, a remote support tool or a management console, and then go for the hypervisor and the backups. We covered two such chains in detail: from a Check Point VPN flaw to ransomware and ransomware against vCenter and ESXi.
Note one detail in the monthly table: no entry added from September onwards carries the "Known" flag yet. That does not mean autumn was free of ransomware. CISA updates the flag when evidence emerges, so the recent months will probably fill in.
Old Vulnerabilities, New Entries
Not everything in the 2026 additions is new. 187 entries have a CVE identifier from 2026 and 24 from 2025, but 39 entries carry a CVE from 2024 or earlier, and 16 from 2020 or earlier. The oldest two date from 2008: a Windows buffer overflow and a Cisco IOS cross-site request forgery flaw. CISA lists them now because it has new evidence of exploitation, which usually means attackers found a population of unpatched systems still online. For a mid-sized company the lesson is simple: the forgotten server, the old router at a branch office and the end-of-life appliance in the plant are exactly what these entries describe.
AI Tools and the Software Supply Chain
Two groups were almost absent from KEV a few years ago and are now a steady presence.
AI and automation platforms: 14 entries. Langflow appears five times (one entry is recorded under IBM), BerriAI LiteLLM three times, and n8n, Marimo, Ray, MLflow, Kestra and GPUStack once each. These tools are often installed by a data or development team, exposed to make a demo work and never added to the asset inventory. They also tend to hold API keys for paid AI services and cloud accounts. Our article on AI infrastructure exposed on the internet goes through the cases.
Developer tools and supply chain: 14 entries. Five are compromises of software distribution rather than classic bugs: the Trivy security scanner (added 26 March), the Nx Console extension and TanStack packages (both 27 May, both flagged "Known" ransomware use), Daemon Tools Lite (27 May) and the TrueConf client updater (2 April). In each case a trusted package or update delivered malicious code. The other nine are CI/CD and code repository servers: JFrog Artifactory (4), JetBrains TeamCity (2), GitLab (2) and Gitea. A compromised build pipeline hands over every secret it uses, which is why the response is not just "update the package" but "rotate the credentials". See our article on the Trivy compromise.
30 entries in the period carry a CISA note that the flaw affects an open-source component, library or protocol that can be used by different products. For those, checking only the vendor name in your inventory is not enough: you need to know which of your products embed the component.
A New Rhythm: Three Days and a Forensic Check
Entries added from 11 June 2026 onwards refer to CISA Binding Operational Directive 26-04, "Prioritizing Security Updates Based on Risk", which replaced BOD 22-01 and BOD 19-02. The directive sets remediation timelines for US federal agencies based on whether an asset is exposed to the internet, whether the CVE is in KEV, whether exploitation can be automated and how much control it gives an attacker.
In the catalog, the effect is visible. Of the 162 entries added from 11 June to 10 November, 103 are marked for "forensic triage" and all 103 carry a three-day due date. Another 23 have a three-day deadline without the triage flag, and 36 have fourteen days. In practice CISA is telling federal agencies: for most new KEV entries you have three days to fix, and you must also check whether the device was already compromised before the fix.
BOD 26-04 does not apply to European companies. Its logic does. A patch applied after exploitation closes the door but leaves whoever came in. That is the subject of our piece on why patched is not the same as clean, built on the Ivanti Connect Secure and NetScaler cases of October.
What This Means for Your Patch Policy
A European mid-market company does not need to read 250 advisories. It needs a policy that turns the catalog into decisions. Here is the one we suggest, built on the numbers above.
- Know what you expose before the next entry arrives. Over 30% of the period's entries concern products whose job is to face the internet or give remote control. Keep a current list of every firewall, VPN, gateway, remote access tool and management console reachable from outside, with model and version. If a device is not on the list, it will not be on the patch plan.
- Treat "KEV and internet-facing" as an emergency measured in days. CISA gives federal agencies three days for most new entries. You may not match that everywhere, but write down a target (for example 72 hours for exposed KEV entries) and measure against it.
- Patch, then check for compromise. For edge devices, a fix applied after the exploitation date is not the end. Check the vendor's integrity tools and indicators, review new accounts and configuration changes, and rotate the secrets the device held.
- Do not stop at KEV. KEV is a floor, not a ceiling. Use EPSS for the probability of exploitation, read vendor advisories that say "exploited in the wild" before CISA lists them, and check the European picture in ENISA's European Vulnerability Database. Our article on prioritising with KEV, EPSS and SSVC explains how to combine them.
- Look for the old and the forgotten. 39 entries were CVEs from 2024 or earlier. End-of-life devices and servers nobody owns are where they live.
- Include the tools nobody calls infrastructure. AI frameworks, CI/CD servers, CMS plugins and help desk software made up a large share of the list. They belong in the inventory and in the patch cycle.
- Keep the evidence. For NIS2 entities, Article 21(2)(e) requires vulnerability handling and disclosure. A dated record of what was found, when it was fixed and how the fix was verified is the evidence a supervisor will ask for. See our vulnerability management lifecycle guide.
Latest Entries Worth a Look
Since this analysis stops at catalog version 2026.11.10, three recent entries deserve a mention. On 6 November CISA added a SonicWall SonicOS SSL VPN flaw (CVE-2026-19868) that can let a user with valid credentials for another authentication method bypass SAML controls and the MFA they enforce. On 9 November it added an OpenSSH flaw on Linux (CVE-2026-123158) that could allow authentication bypass. On 10 November it added two Barracuda flaws that can be chained on the Email Security Gateway (CVE-2026-116606 and CVE-2026-111560), covered in our email gateway article.
Checklist: Ten Minutes with the KEV Catalog
- Subscribe to KEV updates (CISA offers email, RSS and the JSON feed) and route them to whoever owns patching.
- Match the vendors and products in the catalog against your asset inventory every week, not every quarter.
- Flag every match on an internet-facing asset as urgent, with an owner and a deadline.
- For edge devices, add a compromise check to the ticket, not just the patch.
- For supply chain entries, list the secrets the affected tool could read and rotate them.
- Re-test after the fix and keep the evidence with a date.
How Orizon Helps
Matching a catalog against an inventory only works if the inventory is complete. RECON External maps what your domains expose to the internet from the domain name alone, matches the detected products to known CVEs weighed by CISA KEV, Exploit-DB, Metasploit and EPSS, and then tests each finding: Proven, To verify or Closed, with the evidence of every test. That gives you the "KEV and internet-facing" list from step 2 without guessing. If you want a one-off picture first, start from our vulnerability assessment. And when you need to prove what an attacker could actually do on a perimeter you authorise, Fireline runs an autonomous penetration test with evidence and a proof-of-fix retest.