What Is at Stake
NGINX handles web traffic and sends requests to the right apps. Around 225 million websites use it. The two critical bugs are tracked as CVE-2026-123633 and CVE-2026-123007. Both carry a 9.2 severity score and only affect servers with HTTP/2 turned on. One of them goes back to NGINX version 1.9.5, released in 2015. Attackers who succeed could steal sensitive data or shut down services.
How the Bugs Work
Wiz researchers said both issues involve memory handling in HTTP/2 traffic. In one case, memory is freed while a request still needs it. In the other, a counting error in flow control leads to a bad memory copy. Rapid7 has published a proof of concept for the second bug. However, Rapid7 noted that a real attack of this kind needs a huge amount of traffic, about 128 GB of HTTP/2 data.
How Fast Fixes Came
Things moved quickly. F5 sent the first bug fix for review on Nov. 10. Wiz flagged another bug on Nov. 16. By Nov. 18, fixes for all 10 issues had been released publicly.
AI Is Speeding Up Bug Hunting
Five research teams found the bugs on their own: Wiz, Ikotas Labs, Metnir, Rapid7 and Xint. Most used AI tools. Xint said one researcher found the first flaw, and then an automated scan found four more a few days later with no human guidance. Xint also showed that a Linux kernel bug can be chained with the NGINX flaws to gain root access on the host.
What Admins Should Do
F5 says it has seen no signs of active attacks so far. Even so, admins should update right away. The fixes are in NGINX 1.30.7 for the stable branch and 1.31.9 for the mainline branch. If a quick update is not possible, turning off HTTP/2 can lower the risk. Some Linux distributions may take longer to ship patched packages.
A Bigger Trend
This is the latest in a string of AI-found flaws in core internet software, including PHP, Apache and Java. Wiz's Tomer Schwartz said teams must "treat every internet-connected system as potentially vulnerable." He added that AI is finding bugs faster than defenders can patch, though he expects it to make software stronger in the long run. Security teams are already struggling with a nonstop flood of new CVEs to patch.
