Linux 7.1 lands with Torvalds calling AI-fueled kernel patches the new normal

Linus Torvalds used the Linux 7.1-rc3 announcement to confirm what he had been hedging on for months: the wave of incoming patches generated with help from AI tools is no longer a temporary spike. It is the new floor for the kernel that runs most Android phones, the majority of the web servers online, and every machine on the Top 500 supercomputer list. The release candidate carries about a third of its patches in the networking subsystem, fixes for use-after-free vulnerabilities in Bluetooth and graphics drivers, and continued progress on the rewritten NTFS driver from Namjae Jeon — which now supports full write operations, delayed allocation, and integration with the iomap framework.

The numbers tell the practical story. Linux 7.0 brought a patch surge so large that Torvalds initially treated it as an anomaly. Two release cycles later the volume has not dropped. It has settled. For maintainers, this means reading and reviewing significantly more code than the previous generation of kernel work assumed. The people running the review process are still human, and the bottleneck has shifted from generation to verification. The code may be written with AI assistance, but every patch that lands in the mainline has been read, criticized, and approved by a maintainer with no AI in the loop.

This is not the same as saying the kernel is now written by AI. Torvalds and the senior maintainers have been explicit on that distinction for two cycles. The rules laid down in 7.0 require contributors to disclose AI involvement, to take responsibility for what they submit, and to understand the code they are sending in. The volume is going up; the standards are not coming down. What is changing is the practical experience of working on Linux: maintainers are spending more time reviewing, contributors are turning around revisions faster, and the gap between proposing a fix and having it land is closing for routine work.

The skeptical reading of all this is harder to dismiss than it would have been a year ago. More code arriving faster means more opportunities for subtle errors to slip through, and the use-after-free fixes in 7.1-rc3 are exactly the class of bug that AI-assisted patches have historically struggled to spot. Memory-safety failures in core subsystems like Bluetooth and the graphics stack are not theoretical — they are the cracks that real-world security advisories exploit. Bigger patch volume only works if review keeps pace. Whether it does is the open question for the next several releases.

The Linux 7.1 stable kernel is expected to ship on June 7, 2026, if the cycle wraps at RC7, or June 14, 2026 if a final RC8 is needed.