The Linux kernel community, led by Greg KH, has introduced a policy rejecting AI-generated patches in the driver staging area to preserve its role as a learning environment for new developers, despite Linus Torvalds’ overall positive stance on AI usage. This targeted restriction aims to maintain educational value and code quality by preventing AI from undermining hands-on skill development and reducing low-quality, unverified submissions.
A new policy has been introduced in the Linux kernel development community specifically targeting the use of AI and large language models (LLMs) in the driver staging subsystem. Greg Crow-Hartman (Greg KH), a key Linux maintainer, announced that patches generated by AI tools for the driver staging area will be rejected by default. This move comes shortly after Linus Torvalds, the creator of Linux, expressed a generally positive stance on AI, emphasizing that Linux is not against AI usage and encouraging contributors to use AI tools if they wish. However, Greg KH’s policy highlights a nuanced approach, restricting AI-generated contributions in a specific part of the kernel to preserve its educational purpose.
The driver staging area in the Linux kernel serves as a training ground for new developers. It contains drivers that are not yet ready for mainstream inclusion and is designed to help newcomers learn the patch submission, review, and testing processes. Greg KH stresses that this area is intentionally left imperfect, with coding style issues and incomplete features, to provide learning opportunities. Using AI to automatically fix or improve this code would undermine the very purpose of the staging area, which is to help developers build their skills through hands-on experience and feedback from maintainers.
Greg KH also addressed the recent flood of AI-generated security vulnerability reports, many of which were duplicates, unverified, or irrelevant. Linus Torvalds had previously warned contributors against submitting raw AI-generated bug reports without proper verification, as many issues flagged by AI tools have already been identified. To maintain quality and prevent maintainers from being overwhelmed, Greg KH requires that any AI-assisted security fixes in the driver staging area must be thoroughly tested on real hardware and clearly documented before being accepted. This ensures that contributions are both valid and beneficial.
The policy does not reject AI usage across the entire Linux kernel but specifically targets the driver staging subsystem to protect its role as a learning environment. Greg KH’s stance aligns with a broader philosophy shared by security expert Bruce Schneier, who distinguishes between “work tasks” and “gym tasks.” Work tasks prioritize efficiency and results, where AI can be a helpful tool, while gym tasks focus on skill development through practice. The driver staging area is likened to a gym where new developers build their abilities, making AI-generated patches inappropriate in this context.
In summary, Linux remains open to AI-assisted development, but with careful restrictions to preserve educational value and maintain code quality. Linus Torvalds supports AI as a useful tool, while Greg KH enforces limits on its use in specific subsystems to prevent undermining the learning process and to reduce low-quality submissions. This balanced approach reflects the community’s effort to integrate AI responsibly, ensuring it complements rather than replaces human learning and expertise in kernel development.