Five teams independently found two critical nginx flaws, and Xint's automated scan found more
Wiz says five teams found the same nginx bugs on their own, and nginx's advisory list went from two in 2025 to 26 this year.
On November 18, 2026, the security company Xint disclosed five flaws in nginx, the web server that sits in front of a huge share of the internet's sites, each of which lets an attacker run code on the server without logging in. F5, which makes nginx, shipped the fixes the same day in nginx 1.30.7 and 1.31.9.
Xint says one of its researchers found the first flaw while preparing for ZeroDay Cloud 2026. A few days later, in Xint's words, "a fully automated Xint scan found the other four without human guidance." Wiz, a second security company, says its own AI tooling found two of the same bugs, and that five teams found those two independently.
- Flaws
- Five nginx remote code execution bugs, plus a Linux kernel flaw
- Found by
- Wiz Research, Ikotas Labs, Metnir, Rapid7 and Xint
- Fixed in
- nginx 1.30.7 stable and 1.31.9 mainline
- Severity
- CVSS 4.0 score of 9.2
What Xint found
Xint calls the set nginx2root. Each nginx bug gets an attacker code execution with the nginx worker's privileges, under a different configuration. Xint chained them with a separate bug in the Linux kernel's SCTP subsystem to get root on the machine, and in its tests the chain also escaped a Docker container running nginx.
The worst is CVE-2026-123007, which Xint says affects every official release from 1.9.5 through 1.31.8. That's a decade of nginx. Brian Pak posted the disclosure for Xint that morning.
Pre-auth RCE in nginx is the bug every hacker dreams of finding. It almost never happens. Today we’re disclosing nginx2root: five nginx vulnerabilities, and an exploit that takes an unauthenticated attacker all the way to root. One of themhits the default config. If you run nginx, patch now. Then read: https://t.co/rDQQAimLZP CVE-2026-123007, CVE-2026-123634, CVE-2026-123633, CVE-2026-121810, CVE-2026-113425
Xint's post doesn't say which AI models its scan used, how long it ran or what it cost. A Forbes report gave those details, but they aren't on any page Xint or Pak published, so we've left them out.
Five teams, the same bugs
The second half of the story is the overlap. Wiz's write-up says Wiz Research used Wiz Atlas, the harness behind its AI code scanning, to find the two HTTP/2 flaws, then used Gemini 4 Carbon and ChatGPT 5.6 Sol to confirm each could be exploited. When Wiz reported the first one, F5 told it another team had already reported it.
- Wiz reports the use-after-free to F5
- Wiz reports the second HTTP/2 flaw
- F5 releases the fixes
Xint's own timeline puts its report to F5, through HackerOne, on October 16, five days before Wiz's. Wiz connects the pattern to Java4Shell, which Calif disclosed a day earlier.
Complexity can no longer be relied upon as a barrier to discovery or exploitation.
Xint put a number on how fast this is moving. By its count, nginx's security advisory list recorded two new vulnerabilities in all of 2025, and 26 this year by November 17, before nginx2root added five more. I'd read that as the clearest sign in either post that something has changed, more than any single bug.
What it means if you run a website
Wiz says nginx powers more than 225 million websites, and that these flaws are in nginx itself. A site with perfectly secure application code can still be exposed. Wiz showed code execution on a server doing nothing but serving ordinary static files, with no backend at all.
Xint looked at exposure through Shodan on November 10. Of 56.7 million hosts running nginx, 11.94 million showed a version number, and 98.9% of those were in the range hit by CVE-2026-123007. Xint says that's a measure of how common the versions are. How many of those servers could actually be exploited is a separate question, and it doesn't answer it.
The fix is to upgrade. If you can't yet, Wiz's advice for the two HTTP/2 bugs is to turn HTTP/2 off until you've patched. Xint gives the same temporary fix for those two, and it does nothing for the other three, which need the upgrade.
Defenders and attackers get the same tools, and that's why the shorter clock matters. Anthropic cut its disclosure deadline to 30 days on November 16 for the same reason. Wiz says it won't release its exploits at this time.
More on Security
- Calif says one researcher using OpenAI's Codex found a critical Java flaw in an hourNovember 18, 2026
- Google finds attackers using an AI tool that turns security patches into exploits, no human involvedNovember 16, 2026
- Anthropic cuts its bug disclosure deadline to 30 days, saying another AI will find the bugNovember 16, 2026
- Justice Department seizes North Korea's FaceStealer servers, saying fake video calls stole over $400 millionNovember 13, 2026