AI is ready to detect more than 2,000 vulnerabilities in every Linux core version

The Chinese artificial intelligence model Z.ai GLM-5.3 has helped identify more than a thousand critical vulnerabilities in large open source projects. Thanks to artificial intelligence, the traffic of bugs found and fixed in the Linux core has reached 2,000 bugs per version – in recent years, this number has not exceeded 500.

Photo credit: Milad Fakurian / unsplash.com

Posted by Greg Kroah-Hartman, Linux Core Developer journeyshows that the number of vulnerabilities discovered in the project has increased dramatically over the past few stable releases – between versions 6.9 and 6.19, approximately 500 vulnerabilities were fixed in each release. With the release of Linux 7.0, this number reached approximately 1,000 vulnerabilities per version, and Linux 7.2 exceeded 1,500. Most bugs in recent releases are detected by AI services, and with the release of Linux 7.3, this number is likely to exceed 2,000.

Image source: kernel.org

At first glance, the figure of 2,000 vulnerabilities per version seems shocking, since a large number of devices run on the Linux core; on the other hand, the core source code contains approximately 40 million lines of code, which means that there may be thousands of vulnerabilities that have not yet been discovered. Most of these are lower priority – they are related to outdated drivers or features that are long out of date. It’s even a good thing for the core: a whole batch of old drivers and an entire ISDN subsystem with dozens of vulnerabilities were removed this year.

As early as May, with the release of Linux 7.1-rc4, this problem attracted the attention of project founder Linus Torvalds; two years ago, the number of calls was 2-3 times a week, and in 2026 it reached 5-6 times a day.