AI Bug Hunting Pushes 2026 CVE Count Past 100,000, Double Last Year
VulnCheck counted 100,222 CVEs in 2026, more than double 2025, with AI-driven bug hunting driving most of the surge. Only 1.4% have been exploited.

Why it matters
- VulnCheck counted 100,222 CVEs in 2026 as of late Friday, more than double the 48,150 published in all of 2025.
- VulnCheck's Josh Bressers: AI-driven vulnerability hunting is driving a majority of the new findings.
- About 1.4% of 2026 CVEs (1,363) are in VulnCheck's Known Exploited Vulnerabilities catalog.
- Median time from CVE publication to known exploitation fell from 120 days in 2025 to 80 days in H1 2026.
- CISA's Binding Operational Directive 26-04, issued June 10, replaced flat patch deadlines with risk-based questions.
The world has logged more than 100,000 security vulnerabilities so far this year — more than double all of 2025 — and VulnCheck says AI-driven bug hunting is behind most of them.
VulnCheck, itself a CVE Numbering Authority, counted 100,222 CVEs as of late Friday morning Eastern time. Josh Bressers, the company's director of coordinated vulnerability disclosure, told The Register: "There's no denying the fact that vulnerability hunting with AI is driving a majority of what we're seeing."
For context, that figure matters because it marks a structural break in how the security industry does its work. Machine-driven discovery is flooding a triage system built for human-scale reporting volumes, and it's forcing vendors, maintainers and federal regulators to rethink what a vulnerability alert actually means.
How fast did the count climb?
The curve bent sharply upward as the year went on. CVE.icu, an open tracker built on NVD and CVE Program data, logged 4,302 new CVEs in January and 17,051 in October. That is roughly a fourfold increase in monthly volume inside a single year.
The tracker's busiest day of 2026 was October 13, with 2,303 CVEs published in 24 hours. All of 2025 produced 48,150 on the same count. CVE.icu's 2026 total stood at 97,773 as of 10:16 a.m. ET Sunday.
That number runs below VulnCheck's 100,222, a gap that reflects different counting methods and cutoff times. By either measure, 2026 has already more than doubled 2025.
Big vendors are part of the surge. Microsoft fixed more than 1,000 flaws in each of its last two monthly Patch Tuesday releases. Its November rollout contained only one bug known to be exploited in the wild, and that flaw was found with help from Claude.
The longer arc is steep, too. Annual CVE totals have risen from 14,642 in 2017 to 48,150 in 2025, per the CVE.icu chart built on NVD and CVE Program records — and 2026 blew past that entire range before the year ended.
Are attackers actually using more of these bugs?
No — and that's the twist in the story. About 1.4% of this year's CVEs, or 1,363 vulnerabilities, have landed in VulnCheck's Known Exploited Vulnerabilities catalog, Bressers said. He added an important caveat: 1% of 100,000 is far more than 1% of 1,000.
VulnCheck's first-half exploitation report, published in July, found that the ratio of newly exploited bugs to newly published ones peaked at 2.7% in the second half of 2023 and fell to 1.4% in the first half of 2026.
The AI-discovered bugs don't get exploited at a higher rate, either. Of 1,061 flaws VulnCheck attributed to AI-assisted discovery in the first six months of 2026, 14 had been confirmed as exploited. That's 1.3%, close to the rate for everything else.
How fast do attackers move once a bug goes public?
Faster than ever. VulnCheck measured the median time from CVE publication to known exploitation at 120 days in 2025 and 80 days in the first half of 2026.
A disclosure VulnCheck published Friday shows how compressed that window can get. The company reported 14 flaws in Copeland's XWEB EVO refrigeration supervisors. One of them had no public exploit code — only a brief NVD entry that went live at 3:18 p.m. ET on Thursday.
VulnCheck's sensors logged attacks on it one hour and 40 minutes later.
Why does Bressers call this 'a data problem'?
Because the industry now treats every critical score as equally urgent, and the math no longer supports that. Bressers argues defenders need to prioritize by actual risk to their environment rather than severity labels.
Federal agencies already have to. CISA's Binding Operational Directive 26-04, issued June 10, replaced flat patch deadlines with four questions:
- Whether the asset is publicly exposed
- Whether the bug is in CISA's KEV catalog
- Whether an attacker can automate the exploit
- Whether exploitation gives partial or total control
"It's telling people to figure out the vulnerabilities that actually affect you, and then put those at the top of the list, instead of just saying we're going to fix all the criticals," Bressers said.
How are open-source maintainers coping?
By changing their intake rules. The curl project began handling low- and medium-severity reports in public GitHub issues on October 26, keeping only high and critical ones private on HackerOne.
Lead developer Daniel Stenberg wrote that the change "is entirely triggered by the current avalanche of vulnerability reports that keep pouring in." He noted that over 90% of submissions in recent years rated below high severity.
AI labs, meanwhile, are selling defenders the same class of tools driving the surge. OpenAI launched its Daybreak products on Thursday — agents that test and patch code.
What happens next?
Bressers doesn't expect the count to level off. "While 100K sounds like a lot, in a couple years we're going to say, remember when it was only 100K?" he said.
If the exploitation ratios hold while volumes keep climbing, the industry's challenge shifts from finding bugs to filtering them — a problem CISA has already codified for federal agencies and that projects like curl are solving in public, one low-severity report at a time.
Source: webpronews.com
More from Elena Vasquez
Show full bio
Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.
590 articles
Related articles
- Microsoft Patches 1,093 Flaws, Including Outlook Zero-Day Found With Claude
- Six Fake SQLite CVEs and a 10x Surge Expose a Broken Pipeline
- 85% of Security Leaders Report Patching-Delay Incidents, Azul Survey Finds
- METR audit: Anthropic's AI progress may have doubled in speed by July
- Australia's cyber attacks jump 39% as ransomware hits record highs