Greg Kroah-Hartman Bans AI Code in Linux Staging Area

Maintainer Greg Kroah-Hartman officially locked the doors of the drivers/staging subsystem to patches generated by artificial intelligence. The staging folder serves as a digital gymnasium where junior programmers learn the craft of kernel development by fixing minor bugs in incomplete hardware drivers. Automatically generated cosmetic scripts were ruining that learning curve by turning a training ground into a dumping site for machine fluff.

AI-generated contributions will now be rejected by default unless they address a verified security flaw. Even then, submitters must test the code on physical hardware instead of virtual simulators and submit an explicit hardware test report. Machine learning models currently hallucinate or introduce broken code in roughly a third of all security patches, making human validation mandatory.

The policy update comes as automated vulnerability scanners recently flooded the kernel mailing lists with 432 vulnerability disclosures in a single day. The project registered 2,308 CVE records in the first half of 2026, pushing Linux to the top spot among all software vendors for reported security flaws.

The open-source ecosystem is now confronting decades of accumulated technical debt accelerated by algorithms that write code faster than humans can review it. When machines generate bugs and other machines output broken patches, human maintainers remain the only barrier between working infrastructure and total software collapse.

Source: kernel.org