Linus Torvalds says the flood of AI-generated Linux kernel bug reports may finally be showing signs of slowing, although he cautioned that the apparent improvement is limited to some parts of the codebase. In his announcement of Linux 7.4-rc2, the second release candidate for the next major kernel version, Torvalds said the unusually large update included “various AI fixes.”
The release candidate follows what Torvalds described as the largest -rc1 in kernel history. After accounting for automated Coccinelle changes spread across thousands of files, the new work is distributed broadly across graphics, networking, eBPF, filesystems, memory management and architecture code. Linux 7.4-rc2 is a test snapshot rather than a final production release, so the changes remain subject to review and regression testing before the stable kernel is published.
The clearest positive signal came from eBPF, the Linux kernel facility that allows sandboxed programs to run in kernel space. Torvalds wrote that eBPF co-creator and maintainer Alexei Starovoitov claimed fixes in the subsystem were arriving faster than new reports. “That sounds lovely if true,” Torvalds wrote, while adding that most other kernel areas had not reached the same point and urging readers not to assume the broader wave had ended.
The trend follows Anthropic’s April announcement of Mythos Preview, an AI system the company said could identify high-severity vulnerabilities across major operating systems and browsers. More capable automated analysis can expose defects that conventional review misses, but it also creates a second challenge: maintainers must triage, reproduce and safely fix a larger volume of reports without introducing regressions.
Neither Torvalds’ release note nor the 7.4-rc2 announcement identifies a specific CVE, CVSS score, public proof of concept or exploitation-in-the-wild claim tied to this patch wave. Administrators should therefore treat the release candidate as evidence of ongoing development activity, not as a security bulletin or a reason to deploy untested code. Kernel teams should continue monitoring stable-kernel advisories and vendor backports, prioritizing fixes only after confirming the affected versions and operational impact.
