Chalk up a new record for vulns. The number of Common Vulnerabilities and Exposures (CVEs) found and disclosed so far in 2026 surpassed 100,000 on Friday - more than double 2025’s total - and the reason can be summed up in two letters: AI.
“It’s kind of wild,” Josh Bressers, VulnCheck’s director of coordinated vulnerability disclosure, told The Register. “The thing that almost scares the bejesus out of me is while 100K sounds like a lot, in a couple years we're going to say, remember when it was only 100K?”
As of the time of our interview, about 4 pm UTC on Friday, VulnCheck counted 100,222 CVEs, he said. Microsoft alone added 1,094 security flaws to the total count on Tuesday, following 1,474 vulnerabilities the month before.
“There's no denying the fact that vulnerability hunting with AI is driving a majority of what we're seeing.” All the major CVE Numbering Authorities (CNAs) - these are the organizations authorized by the CVE program to assign unique identifiers to security vulnerabilities - like VulnCheck, Microsoft, Oracle, GitHub, and other large enterprise IT vendors “are just screaming into the atmosphere right now,” Bressers said. “It’s AI.”
Meanwhile, the percentage of these bugs that are exploited in the wild remains relatively consistent. About 1.4 percent (1,363) of the CVEs published this year have been added to VulnCheck’s Known Exploited Vulnerabilities (KEV) catalog.
“But obviously, as the number of CVEs grows, [1 percent] of 1,000 is a lot different than [1 percent] of 100,000,” Bressers said.
Plus, as other security researchers have documented, the window between the public release of a security patch and attackers developing working exploits – often by reverse engineering the patch while end users are still applying the fix – is rapidly shrinking. This is also thanks to AI.
From CVE to KEV
From January to June, 88 percent of the exploitation observed by CrowdStrike using public proof-of-concept (PoC) code occurred within 48 hours of the code's release. The security firm said China-linked groups such as Vault Panda and Genesis Panda moved even faster, launching attacks within 24 hours of disclosure. Similarly, VulnCheck found the median time from CVE publication to exploitation falling from 120 days in 2025 to 80 days during the first half of 2026. This window has closed even further during the second half of the year, which also seems to coincide with the AI-induced vulnpocalypse.
“It’s often just hours,” Bressers said, noting 14 make-me-root vulnerabilities that VulnCheck disclosed on Friday in Copeland refrigeration controllers. Two are under active exploitation. These are the devices that Iranian hackers abused to spoil food and spark fires in supermarkets across Europe.
“We published those CVEs, and literally minutes later, we were seeing hits on our honey pots,” Bressers said about the Copeland device vulnerabilities. “We’ve hit a point where with anything that looks interesting, someone is taking that information and seeing what they can do with it. Is it AI driven? Probably. It'd be hard to say it's not, because we've seen it really ramp up recently.”
This influx of software bugs requires a sea change in how the entire industry approaches vulnerability disclosure and patching.
“We don't have a security problem anymore,” Bressers said. “We have a data problem now. Instead of obsessing over vulnerabilities, we need to start looking at things like heat maps for infrastructure, and figure out the vulnerabilities that actually affect you. The CISA BOD helps with that.”
What to prioritize in a patchpocalypse
The US Cybersecurity and Infrastructure Security Agency (CISA) in June issued a cybersecurity directive that replaces flat-deadline vulnerability management policies. Instead of prioritizing patching based solely on CVSS severity scores, the directive instructs organizations to prioritize based on specific risk criteria, including whether the asset is internet-exposed, under active exploitation, and whether an attacker can automate the exploit.
“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.
AI-generated vulnerability disclosures also create huge headaches for open source software developers and maintainers, which Bressers describes as a security problem “super close to my heart.”
“You've got individual developers that are drowning in vulnerability reports,” he said. “You have foundations being crushed by AI agents scraping their data or down their packages, millions of times a day, and it's straining the seams absolutely everywhere you look. The one thing I can say with certainty is: The way we currently do things has to change, and it's going to have to change fairly soon.”
Different open source projects have taken different approaches to solving this issue. The Eclipse Foundation, which houses the Open VSX Registry, in April launched a “managed tier” for enterprise customers and other organizations running Open VSX at scale.
Instead of using the free, open source version, the managed version guarantees a 99.95 percent uptime SLA backed by service credits. Its customers include AWS, IBM, Google, Cursor, and Windsurf. This paid option means commercial users who rely on the registry help fund its ongoing maintenance - and their own platforms’ security and performance.
Meanwhile, curl’s maintainer, Daniel Stenberg, citing an “avalanche of vulnerability reports," updated its bug reporting rules last month. As of October 26, all low- and medium-severity reports go directly into the public GitHub repository, with only the high- and critical-flaws receiving private handling.
Stenberg already shuttered curl's bug bounty program this year as AI-generated submissions piled work onto its security team.
“A lot more needs to happen, and I think it is going to be a challenge to enact some of this change,” Bressers said. “When a package registry drastically changes something, people start freaking out, and the world is going to end, and people on LinkedIn are complaining - a lot of people are hesitant. No one wants to be the first to jump into the alligator-filled pool. That's one of our challenges.”®