# How many CVE records will NIST's National Vulnerability Database list as published from 1 January to 31 March 2027?

> As of 13 November 2026, Prolepsia’s median forecast is 69,500, with an 80% chance it falls between 42,900 and 114,300.

- Forecaster: Prolepsia (https://prolepsia.com)
- Forecast made: 2026-11-13
- Question type: Number
- Resolves: 2027-04-15
- Status: Open
- Forecasts: 1
- Page: https://prolepsia.com/forecasts/barometer-tide-fulcrum/how-many-cve-records-will-nist-s-national-vulnerability-database-list-as-publish
- Open in the app: https://app.prolepsia.com/library/490d590c-0bcc-431b-a944-e1a48634f5b5

## Forecast

- 10th percentile: 42,900
- Median: 69,500
- 90th percentile: 114,300

## How Prolepsia reasons

## TL;DR
The forecast for 1 January–31 March 2027 is a median (50th percentile) of 69,500 CVE records, a 10th percentile of 42,900 CVE records, and a 90th percentile of 114,300 CVE records. Use the median as the central planning figure and the percentile range to stress-test capacity. The forecast puts continued expansion at the center; the decisive risk is that the recent publication plateau persists.

## Context
This question counts non-rejected records with NVD publication timestamps during the target quarter, regardless of analysis status or the year in their CVE identifiers. Resolution uses the API’s `totalResults`, queried on 15 April 2027 or within seven days after. It does not count discoveries, patches, reserved identifiers, or completed analyses ([NVD API documentation](https://nvd.nist.gov/developers/vulnerabilities)).

As of November 13, 2026, publication volumes have moved far above the early-year baseline. The question is no longer simply whether that higher level exists. It is whether publication capacity keeps expanding through the target quarter. NIST’s decision to prioritize enrichment does not itself constrain this count: submitted CVEs remain listed even when they are not selected for immediate analysis ([NIST](https://www.nist.gov/news-events/news/2026/04/nist-updates-nvd-operations-address-record-cve-growth)).

## Evidence
The historical backbone shows growth, but does not establish a stable seasonal rule. The CVE Program reported 40,077 publications in calendar 2024 and 48,244 in calendar 2025. Its 2026 quarterly sequence was 15,154 in Q1, 20,696 in Q2, and 36,962 in Q3. These are upstream program statistics, not the exact NVD resolution series ([CVE Program metrics](https://www.cve.org/About/Metrics)).

The available year-boundary history also argues against assuming a large January reset. The full sequence of nine transitions is below. These figures combine seasonality with underlying growth; they do not isolate a seasonal effect ([CVE Program metrics](https://www.cve.org/About/Metrics?enkwrd=xerox+)).

| Year boundary | Q4 publications | Following Q1 publications |
|---|---:|---:|
| 2017–2018 | 3,570 | 3,935 |
| 2018–2019 | 3,614 | 3,245 |
| 2019–2020 | 4,822 | 4,807 |
| 2020–2021 | 4,387 | 4,415 |
| 2021–2022 | 5,200 | 6,015 |
| 2022–2023 | 6,231 | 7,015 |
| 2023–2024 | 7,876 | 8,697 |
| 2024–2025 | 11,073 | 12,009 |
| 2025–2026 | 12,796 | 15,154 |

This history supports retaining an elevated baseline into the new year. It does not justify adding a mechanical seasonal bonus. The CVE Program also freezes prior-year statistics after reconciliation, while the resolving NVD count excludes later rejections. That helps explain why closely related series need not match exactly ([CVE Program metrics](https://www.cve.org/About/Metrics)).

The resolution-matched NVD observations show the more important change: a sharp rise within 2026. The following counts use non-rejected records and reflect November 13 observations. Quarterly totals, complete months, and the partial November window are labeled separately ([NVD API](https://services.nvd.nist.gov/rest/json/cves/2.0)).

| Publication window | Non-rejected CVE records |
|---|---:|
| January–March 2025 | 11,946 |
| January–March 2026 | 15,152 |
| June 2026 | 7,942 |
| July 2026 | 9,770 |
| August 2026 | 12,255 |
| September 2026 | 14,928 |
| October 2026 | 17,051 |
| November 1–12, 2026 — partial month | 7,008 |

October alone exceeded the entire first quarter of either preceding comparison year. That makes the older year-over-year quarterly increase a poor guide to the current publication level. The late-year rise is the stronger anchor for this forecast ([NVD API](https://services.nvd.nist.gov/rest/json/cves/2.0)).

The latest observations also supply the strongest counterweight. October averaged 550.03 records per calendar day. The complete October 16–November 12 window contained 15,541 records, or 555.04 per day. November’s first twelve days averaged 584, but release-day batching makes that shorter window less informative than it first appears ([NVD API](https://services.nvd.nist.gov/rest/json/cves/2.0)). **The median requires renewed expansion, not merely persistence of the latest flow.** This is why a plateau remains a serious downside risk despite the earlier surge.

The increase is broad enough to resist a single-publisher explanation. An October extraction identified 250 publishing CVE Numbering Authorities, or CNAs. The largest contributed 2,478 of 17,060 records in that snapshot. Its total differs slightly from the later NVD observation, so it is evidence about breadth rather than a substitute resolving count ([October CNA extraction](https://raw.githubusercontent.com/jgamblin/monthlyCVEStats/main/outputs/2026/October/October.md)). Breadth supports persistence, but does not remove concentration or batch risk.

Operating reports strengthen the case for continued supply. GitHub’s January-to-May private reporting flow rose from approximately 550 to over 3,000 reports weekly. Repository advisories rose from approximately 650 to over 5,000 weekly. May CVE requests approached 4,000, nearly ten times their year-earlier level, while processing delays lengthened from roughly one week to multiple weeks. These overlapping workflow measures cannot be added together. They show both more incoming work and growing pressure on the people processing it ([GitHub](https://github.blog/security/supply-chain-security/inside-the-advisory-database-and-what-happens-when-vulnerability-volume-breaks-records/)).

AI-assisted discovery provides a durable mechanism, but its headline totals overstate immediate publication supply. Microsoft’s Linux study moved from 26,502 raw findings to 59 public reports and 33 CVEs after screening, consolidation, and review. Its reported release delay was 41 days at the median and 108 days at the maximum. This is a selected cohort, not an industry-wide conversion rate, but it shows why discovery and publication can move at different speeds ([Microsoft](https://www.microsoft.com/en-us/security/blog/2026/11/11/lessons-from-59-linux-kernel-bug-reports/)).

Publication capacity is changing too. OpenJS described automatic CVE publication integrated with Node.js releases. Its CNA had issued 177 records during January 1–November 12, 2026, versus three during calendar 2025. Those totals are too small to explain the aggregate surge. Their value is as evidence that manual publication steps are being removed ([OpenJS](https://openjsf.org/blog/nodejs-issues-its-cves-through-openjs-foundation)). Oracle’s move toward monthly security releases provides another example of a changing release cadence, without making patches equivalent to newly published CVEs ([Oracle](https://blogs.oracle.com/security/accelerating-vulnerability-detection-and-response-at-oracle)).

Backfill remains a competing explanation for part of the rise. A selected study of 10,240 CVEs matched to reviewed advisories found 373 assigned at least 90 days after the advisory. Of those, 358 advisories already listed fixes. This demonstrates retrospective assignment, but the selected sample does not establish the backfill share of all publications or the size of the remaining inventory ([Safeguard](https://www.safeguard.sh/resources/blog/old-bug-new-cve-id-late-cve-assignment-2026)).

Policy changes cut both ways. Stronger separate-ID wording can support finer-grained assignment, but the official announcement describes the change as a clarification, not a quantified volume expansion ([CVE Program](https://www.cve.org/Media/News/item/blog/2026/10/20/CNA-Rules-v4-2-1-in-Effect)). The frontier-AI researcher CNA pilot began July 28, 2026, with two organizations and an initial six-month evaluation. Its review falls around the target quarter, but no committed expansion schedule or quantified publication pipeline is established ([pilot documentation](https://github.com/CVEProject/frontier-ai-researcher-cnas-pilot)). Conversely, GitHub’s intake limits and Google’s pause on new OSS VRP product-vulnerability submissions show that higher discovery volume can trigger brakes rather than automatic throughput growth ([GitHub](https://docs.github.com/en/code-security/how-tos/report-and-fix-vulnerabilities/report-privately); [Google](https://github.com/Google/bughunters/commit/f8bf23a)).

Rejections matter because the question measures the later, non-rejected population. An Edgescan study identified 5,115 rejections, but only 1,058 withdrawals of previously published records. In its partially observed first-half 2026 cohort, 107 of 21,337 eligible publications were rejected within 180 days. This weighs against assuming that most recent growth will disappear through rejection. It does not establish the future rejection rate, and the study was not independently replicated ([Edgescan](https://www.edgescan.com/wp-content/uploads/2026/11/False_CVE_Research_PaperJM.pdf)).

## What's non-obvious
The binding constraint is not necessarily where the visible backlog sits. **NVD enrichment and CVE publication are different processes.** A growing analysis backlog does not mechanically cap the resolving count, because unenriched records remain listed. The more relevant constraints lie upstream: validating findings, consolidating duplicates, coordinating fixes, assigning identifiers, and publishing records ([NIST](https://www.nist.gov/news-events/news/2026/04/nist-updates-nvd-operations-address-record-cve-growth)).

The opposite mistake is to treat every discovery as an imminent CVE. Broad reporting growth and new publication channels support a higher sustained level. They do not prove that the summer’s percentage growth continues. The latest four-week flow shows why both statements can be true: publication volume can remain historically high while acceleration stops. The forecast reflects expansion beyond that plateau while preserving substantial uncertainty about whether the expansion arrives.

## Uncertainties
The largest gap is the composition and maturity of the publication pipeline. There is no defensible aggregate split between fresh discovery, retrospective assignment, and accumulated review work. A publisher-level inventory of validated, unpublished findings, with expected disclosure dates, would help distinguish durable supply from a finite backlog.

The second gap is future conversion capacity. Automation can increase throughput, while maintainer overload and tighter intake rules can reduce it. The evidence establishes both mechanisms, but not their net effect during the target quarter. Complete monthly counts and publication output by major CNA would be more informative than short windows around large vendor releases.

Finally, status changes and publication-date corrections can alter the April count. Historical snapshots are informative, not immutable, and no earlier count creates a floor for the future quarter. The main uncertainty remains the publication regime itself: continued expansion, a persistent plateau, or renewed acceleration.

## Forecast history

| Date | Forecast |
|---|---|
| 2026-11-13 | Median 69,500, 80% range 42,900 to 114,300 |

## Sources Prolepsia read

- [NVD API documentation](https://nvd.nist.gov/developers/vulnerabilities)
- [NIST Updates NVD Operations to Address Record CVE Growth \| NIST](https://www.nist.gov/news-events/news/2026/04/nist-updates-nvd-operations-address-record-cve-growth)
- [CVE Program metrics](https://www.cve.org/About/Metrics)
- [Inside the Advisory Database and what happens when vulnerability volume breaks records - The GitHub Blog](https://github.blog/security/supply-chain-security/inside-the-advisory-database-and-what-happens-when-vulnerability-volume-breaks-records)
- [Lessons from 59 Linux Kernel Bug Reports \| Microsoft Security Blog](https://www.microsoft.com/en-us/security/blog/2026/11/11/lessons-from-59-linux-kernel-bug-reports)
- [Node.js now issues its CVEs through the OpenJS Foundation CNA \| OpenJS Foundation](https://openjsf.org/blog/nodejs-issues-its-cves-through-openjs-foundation)
- [Accelerating Vulnerability Detection and Response at Oracle  \| security](https://blogs.oracle.com/security/accelerating-vulnerability-detection-and-response-at-oracle)
- [Old Bug, New CVE: Why 373 CVEs in 2026 Arrived Months After the Advisory](https://www.safeguard.sh/resources/blog/old-bug-new-cve-id-late-cve-assignment-2026)
- [GitHub - CVEProject/frontier-ai-researcher-cnas-pilot: This pilot is intended to help the CVE Program evaluate how the large-scale, AI-enabled vulnerability research capabilities of frontier AI labs can contribute to accurate, coordinated, actionable, and sustainable CVE Records. · GitHub](https://github.com/CVEProject/frontier-ai-researcher-cnas-pilot)
- [Privately reporting a security vulnerability - GitHub Docs](https://docs.github.com/en/code-security/how-tos/report-and-fix-vulnerabilities/report-privately)

## The question

NIST's National Vulnerability Database (NVD) lists every published CVE record. Its count of new records has grown fast: non-rejected records published were 11,946 in January–March 2025 and 15,152 in January–March 2026, and 17,051 in October 2026 alone, according to the NVD API on 13 November 2026. Monthly counts rose from about 7,900 in June 2026 to about 17,000 in October 2026. A public question on the 2027 total exists, but its range tops out at 180,000. This question asks about the first quarter of 2027 with a wider range.

### How it resolves

The question resolves to the totalResults value the NVD CVE API 2.0 (https://services.nvd.nist.gov/rest/json/cves/2.0) returns for pubStartDate=2027-01-01T00:00:00.000Z, pubEndDate=2027-03-31T23:59:59.999Z and the noRejected parameter, queried on 15 April 2027 or within seven days after. If the API is unavailable, use the NVD's own published statistics or data feeds that count the same records, published dates in UTC, rejected records excluded.

### Fine print

The count is of records listed as published in the period, whatever their analysis status, and whatever year their CVE identifier carries. Records that NVD later marks as rejected before the query date are excluded.

---

Prolepsia’s forecasting engine made this forecast: it researched the question, weighed the evidence and wrote the report on this page. A probability is not a promise; this one changes as the evidence does.
