{"id":166425,"date":"2026-03-26T17:13:13","date_gmt":"2026-03-26T15:13:13","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai"},"modified":"2026-03-26T17:13:13","modified_gmt":"2026-03-26T15:13:13","slug":"intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai","title":{"rendered":"Interview with Greg Crow-Hartman about AI-generated error reports","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
At the KubeCon Europe conference, The Register interviewed Greg Kroah-Hartman, who is responsible for maintaining the stable and 'staging' branches of the Linux kernel, serving as the main maintainer in 16 kernel subsystems. The interview discusses Greg's perspective on bug reports generated by AI. AI is already being used in the kernel for reviewing changes in the networking subsystem, eBPF, and DRM, and a tool called Sashiko from Google has recently been integrated to check submitted changes. <\/p>\n
Some of Greg's statements: <\/p>\n
\n
'A few months ago, we were getting what we called AI junk \u2014 AI-generated security reports that were clearly incorrect or of low quality. It was even amusing. We didn't really worry about it... A month ago, something happened, and the situation changed drastically. Now we are getting real reports.'\n
'This situation is not just happening in Linux \u2014 all open source projects are receiving real, AI-generated reports, and they are now of high quality and valid. Security teams in major open projects are informally documenting the same trend in their discussions.'\n
When asked what was causing this, Greg replied, 'We don\u2019t know. It seems that no one knows. Either many tools have gotten significantly better, or people have started saying, \u2018Hey, let\u2019s figure this out.\u2019 It appears this concern is affecting many different groups and companies. As far as the kernel is concerned, we can handle it. Our team has grown significantly, is very distributed around the world, and our growth is a reality that isn\u2019t slowing down. These are minor issues, nothing serious, but assistance may be needed in this area for all open projects. Smaller projects have far less capacity to manage the sudden surge in AI-generated bug and vulnerability reports that mention real bugs rather than junk.'\n
Greg mentioned that when he asked the AI to find errors in the proposed set of changes, it found 60 and provided patches for fixing them. Only a third of the identified errors were actual issues, and only 2\/3 of the patches turned out to be correct solutions that didn't require further refinement. However, all of this was far from useless. According to Greg, maintainers cannot ignore this, especially as the performance of AI continues to improve. To mark the patches created using AI, a tag 'Co-developed:' was added. Despite separate attempts to use AI for creating new functionalities, mainly, AI is still used in the core for reviewing changes.\n
One of the most notable advantages of AI is said to be the reduction in patch processing time. When the AI assistant identifies obvious problems, patch authors receive feedback long before the actual maintainer would have time to read the patch: 'If I see the system responding to something, it gives feedback to the author faster than the maintainer could, and that's great. We already have a number of bots that check patches. If I notice that they're returning an error, I immediately understand that as a maintainer I don't even need to look at it. And the developer thinks, 'Oh, I can create another version tomorrow,' which helps improve feedback slightly.' <\/ul>\n