Linux kernel AI code review has moved past its initial stabilization phase and settled into a healthier operating model, project founder Linus Torvalds told The Register on 1 November 2026. The transition comes as subsystem maintainers continue to inspect machine-assisted contributions that still often get rewritten before reaching mainline releases.
Key points
- Linus Torvalds said the Linux kernel project has moved out of an early stabilization phase for artificial intelligence contributions.
- The kernel community now operates under what Torvalds called a “much healthier model” for evaluating machine-generated patches.
- Code submissions produced by automated tools frequently require complete rewrites by human maintainers prior to merging.
- Strict mailing list review workflows remain the primary safeguard against faulty logic entering the core operating system tree.
Moving past the stabilization phase
Development teams working on the Linux kernel have spent recent cycles adjusting to an influx of machine-authored code. As software generation tools grew common across programming communities, kernel maintainers saw an increase in automated bug fixes, drivers, and diagnostic patches sent to public mailing lists.
Torvalds indicated that the community has left behind the initial friction surrounding these tools. Rather than causing ongoing disruption across developer lists, handling machine-assisted proposals has become part of routine project intake. Torvalds described the updated approach as a much healthier model for the project as a whole.
Why maintainers still rewrite automated patches
A calmer intake workflow has not solved the underlying quality problems found in automated code. Torvalds pointed out that machine-made contributions often get rewritten by human developers before they can pass review.
Kernel software demands precise memory handling, strict locking behaviour, and deep familiarity with hardware interfaces. While coding models can produce syntactically valid C that builds without warnings, they frequently miss low-level edge cases and concurrency rules. Maintainers inspecting those submissions must often rework the logic entirely to preserve system stability.
Accountability under the current review model
The Linux kernel relies on an established hierarchy of subsystem maintainers who inspect changes before passing them up to Torvalds. Contributors must sign off on their submissions, taking personal responsibility for the provenance, licensing, and reliability of the code they submit.
That structure treats machine-assisted submissions no differently from any other contribution. Patches that fail to meet project standards are rejected or revised on the mailing lists, keeping the burden of verification on the developer rather than weakening review standards for automated output.
What happens next for kernel development
The Linux kernel community plans to maintain its standard review tiers through upcoming merge windows. Maintainers will continue to assess pull requests on their engineering merits, relying on human oversight to filter machine-assisted submissions before they enter production kernels.
