Linus Torvalds says AI-reviewed fixes are making larger Linux updates routine
Linux 7.2-rc7 was larger than Torvalds preferred, and he says AI tools are increasing development activity while straining bug triage.
By Renata Fuchs · Policy Reporter
· 3 min read
Linus Torvalds said the Linux 7.2-rc7 development build was larger than he wanted, but did not contain anything he considered especially alarming. In his Linus Torvalds AI Linux updates comments, he described the volume as a new normal, with many fixes arising from review by different AI tools, according to The Register.
The remark is a qualified view of AI’s role in the kernel. Torvalds did not say the changes were all generated by AI, nor did he put a number on the share connected to AI-assisted review. He said the immediate release candidate contained many small fixes across drivers, filesystems, core networking and architecture-specific code, alongside a smaller number of larger changes involving s390/zcrypt, btrfs fixup-worker infrastructure and netfilter ipset.
Release candidates are development builds issued before a stable kernel release. Their patch volume alone does not establish that code quality has improved or that the number of security vulnerabilities has risen.
Why are Linux kernel updates getting larger?
Torvalds had flagged unusually large release candidates earlier in the cycle. On August 2, he called Linux 7.2-rc6 the biggest rc6 in years by commit count, according to The Register. The prior Linux 7.2-rc5 was also described as unusually large. In its report on the rc5 announcement, Phoronix said Torvalds noted that similar size had appeared in the 7.1 cycle, though he attributed more than a third of that particular rc5 patch to accumulated networking work.
In a recorded discussion cited by a YouTube transcript, Torvalds said the prior two releases had carried roughly 20% more commits than releases across many preceding years. He said he first suspected the Linux 7.0 version change had prompted contributors to submit more work, but later concluded that AI tools had improved enough to produce a noticeable increase in development across the project.
That is an observation by Torvalds, rather than an independently audited measurement of AI’s contribution. It also spans separate activities that are often grouped together: developers using AI while writing code, tools reviewing code or identifying defects, and people submitting reports based on automated findings.
What problem does AI create for Linux maintainers?
Torvalds has said he finds AI tools technically useful and interesting, while criticizing the operational burden created by their use. In the recorded discussion, he said the kernel security mailing list had been flooded with duplicate reports from tens or hundreds of people using AI tools, sometimes describing the same underlying issue in different ways.
That shifted work to a small security-list group that had to route reports to maintainers familiar with the relevant subsystem. Techzine similarly reported that Torvalds’s criticism was aimed at submitters forwarding AI output without analysis, context or a proposed fix, rather than at the tools themselves.
Torvalds said AI-discovered bugs should generally be considered public because others may be able to identify them as well. He also cautioned reporters not to publish exploit details. The distinction matters for open-source projects: more automated scrutiny may uncover useful issues, but the value depends on whether contributors can help maintainers assess and resolve them.
This story draws on original reporting from The Register.
