Nexus Void Research
Vulnerability Statistics, CVE Statistics, CISA KEV, EPSS, CVSS, NVD, Time to Exploit, Data Breach Cost, Open Source Malware, AI Code Security, Original Research
Vulnerability Statistics 2026: 60+ Data Points (NVD, KEV, EPSS)
Vulnerability statistics 2026: 99,943 CVEs published to Nov 18, 0.44% confirmed exploited, 5-day median to exploit, plus breach cost and AI code data.
Vulnerability statistics for 2026 show record volume and faster exploitation. The National Vulnerability Database (NVD) published 99,943 non-rejected CVEs between January 1 and November 18, 2026, about double the 49,972 CVE records of all of 2025. Only 0.44% of scored CVEs are confirmed exploited in CISA's KEV catalog, but the exploited ones move fast: the median 2026 CVE in KEV was listed 5 days after disclosure.
This page collects 60+ vulnerability statistics for 2026. The core numbers are our own analysis of primary data pulled on November 18, 2026: the NVD CVE API, the CISA Known Exploited Vulnerabilities (KEV) catalog and the full FIRST EPSS score file. Industry figures from Verizon, Mandiant, IBM, FIRST and others are labelled with their source. We update this page as the data changes.
Key takeaways
99,943 non-rejected CVEs were published in NVD in 2026 through November 18, about 310 a day.
2026 CVE volume is 5.3x the 2020 level (19,222 records).
Only 15% of 2026 CVE records carry a CVSS v3 severity in NVD, against 33% for 2025.
0.44% of all EPSS-scored CVEs (1,782 of 408,395) are in CISA KEV.
CISA added 298 CVEs to KEV in 2026, up from 245 in 2025.
The median exploited 2026 CVE reached KEV 5 days after disclosure; 27% were listed on or before disclosure day.
Vulnerability exploitation was the #1 breach entry point in Verizon's 2026 DBIR at 31%.
The global average data breach cost hit $4.99 million (IBM, 2026); in India it reached a record INR 25.5 crore.
82% of organizations carry security debt, flaws left open for more than a year (Veracode, 2026).
How many CVEs were published in 2026?
99,943 non-rejected CVEs were published in NVD between January 1 and November 18, 2026, or 102,726 records if rejected CVEs are included. That is already about double the 49,972 records published in all of 2025, and the pace is still rising: 42,592 non-rejected CVEs were published in the last 82 days of that window alone, about 520 a day.
Year | CVE records published in NVD | Change vs prior year |
|---|---|---|
2020 | 19,222 | |
2021 | 21,950 | +14% |
2022 | 26,431 | +20% |
2023 | 30,949 | +17% |
2024 | 40,704 | +32% |
2025 | 49,972 | +23% |
2026 (to Nov 18) | 102,726 | +106% |
Counts are NVD records by publication date, including rejected CVEs, so years are comparable. 2026 excluding rejected CVEs: 99,943.
The 2026 total has already passed every forecast. FIRST predicted a median of 59,427 CVEs for 2026 in February, with a 90% confidence interval of 30,012 to 117,673, as reported by Infosecurity Magazine. In June it raised the forecast to about 66,000 after January to April came in above the model, per SecurityBrief, with commentators pointing to AI-assisted vulnerability discovery as a driver.
How severe are 2026 vulnerabilities?
Of the 2026 CVEs that carry a CVSS v3 severity in NVD, about half are High or Critical. The bigger story is how many carry no NVD severity at all: 15,666 of 102,726 records (15%) had a CVSS v3 severity in NVD on November 18, 2026, down from 16,665 of 49,972 (33%) for 2025. Volume has outrun enrichment, so "sort by CVSS" leaves most new CVEs unsorted.
2026 CVEs with a CVSS v3 severity in NVD | CVEs | Share |
|---|---|---|
Critical (9.0 to 10.0) | 2,259 | 14.4% |
High (7.0 to 8.9) | 5,833 | 37.2% |
Medium (4.0 to 6.9) | 7,220 | 46.1% |
Low (0.1 to 3.9) | 354 | 2.3% |
Total with a v3 severity | 15,666 | 15% of all 2026 records |
Across all years, 108,555 of 196,640 NVD CVEs with a CVSS v3 severity (55.2%) are rated High or Critical: 30,882 Critical and 77,673 High. Severity alone cannot set priorities when more than half of everything is "urgent". We explain why in what CVSS misses.
How many vulnerabilities are actually exploited?
Very few, and that is the most useful statistic on this page. CISA's KEV catalog listed 1,782 CVEs with confirmed exploitation on November 17, 2026: 0.44% of the 408,395 CVEs that FIRST EPSS scores. EPSS, the Exploit Prediction Scoring System, estimates the probability of exploitation activity in the next 30 days.
Exploitation signal (n = 408,395 EPSS-scored CVEs) | CVEs | Share |
|---|---|---|
In CISA KEV (confirmed exploited) | 1,782 | 0.44% |
EPSS 50% or higher | 4,388 | 1.1% |
EPSS 10% or higher | 17,558 | 4.3% |
In KEV or EPSS 10% or higher | 17,995 | 4.4% |
EPSS 1% or higher | 149,213 | 36.5% |
KEV additions per year: 311 (from the catalog's launch in November 2021), 555 in 2022, 187 in 2023, 186 in 2024, 245 in 2025 and 298 in 2026 so far.
24.5% of KEV CVEs (437 of 1,782) score below 10% on EPSS, so EPSS alone misses confirmed exploitation.
361 KEV entries (20%) are flagged by CISA as known to be used in ransomware campaigns, 27 of them added in 2026.
70% of 2026 KEV additions (210 of 298) are CVEs published in 2026; 29 date from 2021 or earlier, the oldest from 2008.
For how to turn these signals into a fix order, see how to prioritize vulnerabilities with CVSS, EPSS and KEV.
How fast are vulnerabilities exploited in 2026?
In days. In our time-to-exploit study of 219 CVEs published and confirmed exploited in 2026, the median gap from NVD publication to KEV listing was 5 days and the mean 20.5 days.
27.4% were listed in KEV on or before the day they were published in NVD.
81% were listed within 30 days of publication.
Exploited 2026 CVEs rated Critical had a median of 5 days to KEV; High, 3 days; Medium, 10 days.
12% of exploited 2026 CVEs scored below 7.0, so a "fix High and Critical only" rule would have skipped them.
Mandiant's M-Trends 2026 estimates the mean time to exploit at -7 days for 2025, meaning exploitation on average begins before a patch exists, down from 63 days in 2018 and -1 day in 2024, per Help Net Security.
Which vendors and weaknesses are most exploited in 2026?
Microsoft led 2026 KEV additions with 41 CVEs, followed by Cisco with 21. Code injection (CWE-94), improper authentication (CWE-287) and OS command injection (CWE-78) were the most common weakness types among 2026 KEV additions that list a CWE.
Vendor (2026 KEV additions) | CVEs |
|---|---|
Microsoft | 41 |
Cisco | 21 |
10 | |
Apple | 10 |
Linux | 9 |
Fortinet | 9 |
Ivanti | 8 |
Adobe | 7 |
Citrix | 6 |
Check Point | 5 |
Weakness (CWE) in 2026 KEV additions | CVEs |
|---|---|
CWE-94 Code injection | 20 |
CWE-287 Improper authentication | 19 |
CWE-78 OS command injection | 18 |
CWE-502 Deserialization of untrusted data | 14 |
CWE-22 Path traversal | 14 |
CWE-306 Missing authentication for critical function | 13 |
CWE-918 Server-side request forgery | 11 |
CWE-416 Use after free | 10 |
CWE-787 Out-of-bounds write | 10 |
Across the whole KEV catalog, Microsoft accounts for 391 entries, Cisco 103, Apple 96, Adobe 83 and Google 77. Most of the top 2026 weaknesses are application-layer flaws (injection, authentication, deserialization, path traversal, SSRF) that code review and static analysis are built to catch.
How do breaches start in 2026?
Through vulnerabilities, more than any other way. Verizon's 2026 Data Breach Investigations Report (DBIR) found vulnerability exploitation in 31% of breaches, overtaking stolen credentials as the leading initial access vector for the first time in the report's history, per SecurityInfoWatch.
Mandiant's M-Trends 2026, covering 2025 investigations, agrees:
Exploits were the most common initial infection vector at 32%, first for the sixth year running.
Voice phishing was second at 11% of investigations where a vector could be identified.
Global median dwell time rose to 14 days, from 11 days in 2024.
The median handoff from initial access to a second threat group fell to 22 seconds.
How much does a data breach cost in 2026?
IBM's 2026 Cost of a Data Breach Report, based on 602 organizations breached between March 2025 and February 2026, puts the global average at a record $4.99 million, up 12% year on year, per Infosecurity Magazine.
In India the average reached a record INR 255 million (INR 25.5 crore), up 15.9% from INR 22 crore, with 39,500 records compromised in the average breach, according to IBM India.
Open source and software supply chain vulnerability statistics
Malicious packages now outnumber new CVEs. Sonatype's Q3 2026 Open Source Malware Index logged 149,329 malicious open source packages in the quarter, bringing its total since 2017 to 1,960,846. npm accounted for 89.5% of the quarter's malicious packages, down from 96.6% in Q2. In Q1 2026 Sonatype counted 21,764, so quarterly volume grew about sevenfold in six months.
Most of what a dependency scanner reports is never exploited, but the few that are get used quickly, which is why reachability and exploitation signals matter more for dependencies than raw CVE counts. We cover the response in how to automatically fix vulnerable dependencies and reachability analysis.
AI-generated code vulnerability statistics
AI-written code is not getting safer as fast as it is getting more common. Veracode's 2026 GenAI Code Security Report found AI-generated code passed security checks only 56% of the time, with coding-specific models no more secure than general-purpose ones, per SD Times. Pass rates ranged by language from 63% for Python down to 30% for Java. Reasoning models passed 56% of the time against 51% for non-reasoning models, and model size made little difference.
More on this in AI coding assistant security vulnerabilities.
Security debt and remediation statistics
Most organizations are carrying old, known flaws. Veracode's 2026 State of Software Security report, covering 1.6 million applications, found:
82% of organizations carry security debt (flaws open for more than a year), up from 74% in 2025.
60% carry critical security debt (old flaws that are both severe and exploitable), up from 50%.
The median half-life of a flaw across all scan types is 243 days, down from 252 days.
Those figures were reported by Help Net Security. Our AppSec backlog analysis shows why debt piles up: a "fix every High and Critical" policy puts about six times more CVEs in the urgent queue than a "KEV or EPSS 10%+" policy.
Vulnerability statistics 2026: summary table
Statistic | Value | Source |
|---|---|---|
CVEs published in 2026 (to Nov 18, non-rejected) | 99,943 | NVD, our analysis |
CVE records published in 2025 | 49,972 | NVD |
Growth in CVE records, 2020 to 2026 | 5.3x | NVD, our analysis |
2026 CVEs with a CVSS v3 severity in NVD | 15% | NVD, our analysis |
Share of v3-scored 2026 CVEs rated High or Critical | 51.6% | NVD, our analysis |
CVEs in CISA KEV | 1,782 | CISA KEV |
KEV share of all EPSS-scored CVEs | 0.44% | CISA KEV, FIRST EPSS |
CVEs added to KEV in 2026 | 298 | CISA KEV |
CVEs in KEV or EPSS 10%+ | 4.4% | Our analysis |
KEV CVEs with EPSS below 10% | 24.5% | Our analysis |
Median days from disclosure to KEV, 2026 CVEs | 5 | Our analysis |
2026 exploited CVEs listed on or before disclosure | 27.4% | Our analysis |
Breaches starting with vulnerability exploitation | 31% | Verizon DBIR 2026 |
Mean time to exploit (2025) | -7 days | Mandiant M-Trends 2026 |
Global average breach cost | $4.99M | IBM 2026 |
India average breach cost | INR 25.5 crore | IBM 2026 |
Malicious open source packages, Q3 2026 | 149,329 | Sonatype |
Organizations with security debt | 82% | Veracode SoSS 2026 |
AI-generated code security pass rate | 56% | Veracode GenAI 2026 |
What do these vulnerability statistics mean for your team?
Volume is now unmanageable by severity alone. When roughly 300 CVEs a day are published, most without an NVD severity, and only about 1 in 230 scored CVEs is ever confirmed exploited, the job is no longer "fix everything rated High". It is to find the small set that is exploited or likely to be, check whether your code actually reaches it, and fix that set within days.
That is what ARGUS does. It scans your code, dependencies and secrets, then ranks every finding with a 0 to 100 priority score that blends CVSS, EPSS, CISA KEV and reachability, so an exploited, reachable CVE rises to the top and an unreachable one sinks without disappearing. Each dependency finding is marked reachable, unreachable or unknown with an evidence string. When you click Assign to Agent, ARGUS generates the smallest fix, verifies it with a build and a re-scan, and opens a pull request for you to review and merge; it never merges on its own. Its pull request reviewer can block new high-risk findings before they merge.
Limits: reachability is import-level and covers JS/TS and Python only, so other ecosystems are treated as reachable. Read more about SCA with reachability and AI code review, or scan a repository free.
Methodology and sources
Our analysis (primary data, pulled November 18, 2026):
NVD CVE API 2.0: CVE counts by publication date for each year, using windows of 120 days or less, with and without the noRejected filter; CVSS v3 severity counts using the cvssV3Severity filter. NVD's CVSS v3 coverage changes as records are enriched, so the severity shares will move.
CISA KEV catalog version 2026.11.17 (1,782 CVEs): additions by year, vendor, CWE, CVE year and known ransomware use.
FIRST EPSS score file (model v2026.06.15, score date November 18, 2026, 408,395 CVEs): threshold counts and overlap with KEV.
Time to exploit: our study of 2026 KEV additions matched to NVD publication dates (January 1 to November 13, 2026).
Third-party figures: Verizon 2026 DBIR, Mandiant M-Trends 2026, IBM Cost of a Data Breach 2026, FIRST 2026 Vulnerability Forecast, Sonatype Open Source Malware Index (Q1 and Q3 2026), Veracode 2026 State of Software Security and 2026 GenAI Code Security Report. Each is linked to the coverage we verified the figure against.
Frequently asked questions
How many vulnerabilities were reported in 2026?
The NVD published 99,943 non-rejected CVEs between January 1 and November 18, 2026, or 102,726 records including rejected CVEs. That is about 310 a day and roughly double the 49,972 CVE records published in all of 2025, making 2026 the largest year on record by a wide margin.
How many CVEs are exploited in the wild?
CISA's Known Exploited Vulnerabilities catalog listed 1,782 CVEs on November 17, 2026, which is 0.44% of the 408,395 CVEs scored by FIRST EPSS. CISA added 298 of them in 2026. Exploitation evidence is far rarer than high severity, which covers over half of scored CVEs.
What percentage of vulnerabilities are critical in 2026?
14.4% of 2026 CVEs that carry a CVSS v3 severity in NVD are Critical and 37.2% are High, so about half are High or Critical. Only 15% of 2026 CVE records had a CVSS v3 severity in NVD as of November 18, 2026, so most new CVEs have no NVD severity yet.
How fast are vulnerabilities exploited after disclosure?
Fast. In our study of 219 CVEs published and confirmed exploited in 2026, the median time from NVD publication to CISA KEV listing was 5 days, 27.4% were listed on or before disclosure day and 81% within 30 days. Mandiant estimates the 2025 mean time to exploit at -7 days.
Which vendor has the most exploited vulnerabilities in 2026?
Microsoft, with 41 of the 298 CVEs CISA added to KEV in 2026, followed by Cisco with 21, Google and Apple with 10 each, and Linux and Fortinet with 9 each. Microsoft also leads the whole KEV catalog with 391 entries.
What is the most common way breaches start in 2026?
Vulnerability exploitation. Verizon's 2026 DBIR found it in 31% of breaches, overtaking stolen credentials as the top initial access vector for the first time. Mandiant's M-Trends 2026 also ranks exploits first, at 32% of intrusions with a known vector, for the sixth year running.
What is the average cost of a data breach in 2026?
IBM's 2026 Cost of a Data Breach Report puts the global average at a record $4.99 million, up 12% from the prior year. In India the average reached a record INR 25.5 crore (INR 255 million), up 15.9% year on year.
How many malicious open source packages were found in 2026?
Sonatype logged 149,329 malicious open source packages in Q3 2026 alone, bringing its total since 2017 to 1,960,846. npm accounted for 89.5% of the quarter's malicious packages. In Q1 2026 Sonatype counted 21,764, so quarterly volume grew about sevenfold in six months.
Is AI-generated code secure?
Often not. Veracode's 2026 GenAI Code Security Report found AI-generated code passed security checks 56% of the time, with coding-specific models no safer than general ones. Pass rates ranged from 63% for Python to 30% for Java. Review AI-written code with the same SAST, SCA and secret scanning as human code.
How should teams prioritize so many vulnerabilities?
Start with exploitation evidence, not severity. Fix CVEs in CISA KEV first, then high EPSS scores, then adjust for reachability and internet exposure, using CVSS as a tiebreaker. In our data only 4.4% of scored CVEs are in KEV or have an EPSS score of 10% or higher.
DATA SOURCES
NVD CVE API 2.0: https://nvd.nist.gov/developers/vulnerabilities ; CISA KEV v2026.11.17: https://www.cisa.gov/known-exploited-vulnerabilities-catalog ; FIRST EPSS 2026-11-18: https://www.first.org/epss/ ; Verizon DBIR 2026; Mandiant M-Trends 2026; IBM Cost of a Data Breach 2026; FIRST 2026 forecast; Sonatype Q3 2026; Veracode SoSS and GenAI 2026