A Data-Led Look At AI’s Role In CVEs And Patch Delays
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Data-Led Look At AI’s Role In CVEs And Patch Delays on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

An October 26 analysis of NVD and CVE.org records counted 13,934 new CVEs published from October 1–25, 2026, more than three times the total for all of October 2025. The data does not establish how much AI contributed to discovery; the analysis says only 1.4% of October records explicitly credit an AI lab or security firm. It also reports that 516 of 6,157 vulnerabilities Anthropic says it reported to maintainers were patched to the company’s knowledge.

An analysis published October 26, 2026, counted 13,934 CVE records published from October 1 through 25, more than three times the 4,389 recorded during all of October 2025. The figures show a steep rise in disclosed vulnerabilities, but they do not establish how much artificial intelligence contributed to finding them; the analysis says only 1.4% explicitly credit an AI lab or AI security firm.

The analysis says it queried the National Vulnerability Database (NVD) for records by publication date and cross-checked October entries against CVE.org. It reports that the sources agreed on 13,766 records for October plus 202 records later rejected. The stated 13,934 figure covers October 1–25, while the separate monthly figure includes the later rejected records; those totals should not be treated as identical measures. The author says 98% of October records carry a 2026 CVE identifier, and the median interval between reserving an identifier and publication was 15 days.

Publication volume rose through the summer: the analysis counts 8,001 records in June, 9,919 in July, 12,719 in August and 15,074 in September. It says the daily rate’s month-to-month growth slowed to about 11% in October, from 20% to 28% in the preceding months. The article’s estimate that October could end near 17,000 records is a projection, not a final count.

The rise is concentrated among a handful of organizations, according to the analysis. Microsoft’s October count increased from 180 records in 2025 to 1,455 in 2026; Oracle’s rose from 78 to 1,078, with nearly all of the latter published on October 20. GitHub and VulnCheck also recorded large increases. These comparisons describe published records, not necessarily distinct underlying bugs or the amount of AI-assisted work behind them.

At a glance
reportWhen: Analysis published October 26, 2026; CV…
The developmentA data analysis published October 26 reports a sharp rise in CVE records alongside limited public evidence on AI’s role and a substantial reported gap between vulnerability disclosures and patches.
Thorsten Meyer AI
NVD + CVE.org · data to 25 Oct 2026

Infographic 01 · How big the CVE wave really is

The CVE chart is real. Steep, not parabolic.

New vulnerability records have tripled year over year. A handful of big vendors and clearinghouses drive the jump, and the growth rate cooled in October.

13,934new CVE records, 1–25 Oct 2026
3.2×all of October 2025 (4,389)
88,772in 2026 so far vs. 49,972 in all of 2025
+11%daily rate, Oct vs. Sep: growth is cooling

New CVE records per month

January 2024 – 25 October 2026 · phones show the last 13 months

0
5k
10k
15k
3.6k
4.4k
15.1k
13.9k*
Jan ’24Jul ’24Jan ’25Jul ’25O’25N D J’26F M A M J J A S O’26
2024–20252026September 2026 (record)* October 2026: 1–25 Oct only
+18%Jun vs. May
+20%Jul vs. Jun
+28%Aug vs. Jul
+22%Sep vs. Aug
+11%Oct vs. Sep

Growth of new CVEs per day versus the previous month. Steep, but the acceleration slowed in October: not “parabolic”.

Who publishes the jump

records by CVE Numbering Authority, October vs. October

  • MicrosoftOct 2025 → Oct 2026180 → 1,455
  • OracleOct 2025 → Oct 202678 → 1,078
  • VulnCheckOct 2025 → Oct 2026163 → 2,085
  • GitHub (open-source maintainers)Oct 2025 → Oct 2026276 → 1,159
  • Linux kernelOct 2025 → Oct 2026660 → 925

Microsoft’s October Patch Tuesday fixed a record 1,413 CVEs (SecurityWeek (14 Oct 2026)). The Linux kernel grew far less.

Not a backlog dump

checks on the October 2026 records

  • 98% carry a 2026 ID
  • 15 days median from reserving an ID to publishing it
  • 217 numbering authorities published, vs. 163 in Oct 2025
  • 202 records were later rejected (counted, as NVD does)
  • ~17,000 if October’s pace holds: a projection, not a count

Sources: NVD CVE API 2.0 (counts by publication date, all statuses, queried 26 Oct 2026); CVE.org records (credits field, fetched 26 Oct 2026); SecurityWeek (14 Oct 2026). Own analysis; raw files and queries in the kit.
Powered by Thorsten Meyer AI

Why the Patch Gap Matters

More disclosures can help software maintainers and users identify weaknesses, but a published CVE is not itself a fix. The analysis highlights a reported gap in Anthropic’s vulnerability disclosures: of 6,157 vulnerabilities the company says it reported to open-source maintainers, 516 were patched to its knowledge. That is roughly one in twelve, according to the source. The figure reflects Anthropic’s knowledge and should not be read as a complete, independently verified measure of patch status.

The distinction matters because users face risk while flaws remain unpatched, regardless of whether a person or an AI tool found them. The source also cites reporting by Reuters and claims from NorForge that attackers use inexpensive AI agents against older vulnerabilities. Those reports point to a separate concern: AI may help defenders find flaws, while attackers can also use automation against known weaknesses. The supplied material does not quantify the scale of that activity or show that the October CVE surge caused an increase in successful attacks.

For organizations responsible for software, the practical signal is not the headline count alone. It is whether affected products are identified, fixes are released, and users install them. The analysis presents patching capacity as a potential constraint as discovery and disclosure volumes rise.

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

As an affiliate, we earn on qualifying purchases.

What the CVE Counts Show

A CVE is a public identifier for a disclosed vulnerability. The source explains that records are generally issued through a CVE Numbering Authority, often a vendor, as part of coordinated disclosure. That means the existence of a record reflects a disclosure process; it does not identify every person or tool involved in finding the flaw, and it does not mean attackers played no role in discovering it first.

For October, the analysis found that only 6,346 of 13,934 records named a finder. It counted 200 records explicitly crediting an AI lab, model or AI security firm: 123 associated with Anthropic or Claude, 65 with AI security companies, and 16 with OpenAI. The author describes that as a floor because records may omit attribution. The count cannot establish AI’s full contribution, nor support a claim that most vulnerabilities were found by AI.

The source also cites Anthropic’s dashboard, which as of October 2 listed 29,439 candidate vulnerabilities identified by Claude models. External firms reviewed 6,123 and confirmed 92.7% as real, according to Anthropic; the company said 6,157 had been reported to maintainers. Those are company-reported figures, and candidate findings are not equivalent to published CVEs. Anthropic’s reported findings had resulted in 219 CVE records and 365 GitHub advisories at the time described.

“29,439 candidate vulnerabilities found by Claude models.”

— Anthropic, as described in its vulnerability dashboard

What the Records Cannot Establish

The CVE data does not show how much AI contributed to the broader increase. More than half of the October records named no finder, and the attribution fields that do exist may be incomplete. The 1.4% explicitly attributed to AI-related organizations is therefore a documented minimum in this dataset, not a reliable estimate of AI’s actual share.

Patch status is also incomplete: Anthropic’s count reflects what the company knew, and the supplied source does not provide an independent, system-wide audit of whether maintainers fixed the reported vulnerabilities. Nor does the material provide enough detail to compare exploitation rates across years or establish that the CVE publication rise corresponds to a similar increase in attacks. The source notes that CISA’s Known Exploited Vulnerabilities catalog grew less quickly than CVE publications, but the supplied excerpt does not include the full comparison needed to quantify that gap.

Finally, October’s figures are not a finalized month-end total. The analysis’s estimate of nearly 17,000 records assumes the observed pace continues. That estimate could change, and the source does not establish whether the unusually high counts will persist.

Tracking Disclosures and Fixes

The next useful checks are the finalized October CVE count, later updates to attribution records, and whether maintainers publish patches for the vulnerabilities reported by AI security programs. Organizations can also compare published vulnerabilities with vendor advisories and their own asset inventories rather than treating a CVE total as a direct measure of exposure.

For the AI question, better disclosure of tools and contributors would make future counts more informative. For the risk question, updated patch and exploitation data will help show whether the growing discovery pipeline is being matched by remediation. The October analysis offers an early snapshot; it does not resolve either issue.

Key Questions

How many CVE records were published in October through the 25th?

The analysis counted 13,934 records from October 1 through October 25, 2026. It separately reports 13,766 October records plus 202 later rejected, so the figures reflect different record treatments.

Does the increase prove AI is finding most vulnerabilities?

No. The data does not establish AI’s overall contribution. Only 200 records, or 1.4% of the October count explicitly credited an AI lab or security firm, while 54% named no finder, according to the analysis.

How many Anthropic-reported vulnerabilities were patched?

Anthropic said 516 of 6,157 vulnerabilities it reported to maintainers were patched to its knowledge. This is the company’s reported status, not an independent audit of every disclosure.

Does a CVE publication mean a vulnerability is fixed?

No. A CVE is a public identifier for a disclosed vulnerability. The record does not by itself confirm that a patch exists or has been installed.

Is the October total final?

No. The 13,934 count covers October 1–25. The source’s estimate of nearly 17,000 for the full month is a projection, not a confirmed total.

Source: ThorstenMeyerAI.com

You May Also Like

Elon Musk’s SpaceXAI Unveils Grok 4.6, A Game-Changer In AI Performance

SpaceXAI’s Grok 4.6 is announced as a new AI model claiming performance comparable to Fable 5 at a significantly lower cost, but key details remain unverified.

Comparing Apple’s SpeechAnalyzer API With Whisper: Which Reigns Supreme?

An analysis comparing Apple’s new SpeechAnalyzer API with Whisper to determine which offers superior speech recognition for small software teams.

Google’s AI Chip Push: Four New Chips In A Year And A New Nvidia Option

Google announced TPU 8tx for confidential inference and TPU 8i Max for long inputs, bringing its eighth-generation TPU systems to four.

Could One Of These 9 Best External GPUs For 2027 Suit Your AI Projects?

A 2027 buyer’s guide compares nine external GPU options by connection bandwidth, upgradeability, power, cooling and compatibility.