Hi all, I've noticed an increasing number of bug reports submitted to the Debian kernel mailing list that appear to have been generated by LLMs.
A recurring problem is that some of these reports omit the system information required for triage, although the necessary information is already described in the reportbug for debian kernel packages. For example, reports may lack kernel logs, hardware information, or reproduction steps, and may instead include hallucinated diagnostic details. This requires maintainers to request the missing information manually and increases the effort needed to handle the reports. I suspect this problem is not limited to Debian kernel. As mentioned in my previous email, I am therefore planning to propose additional guidance for the Debian kernel wiki and bug-reporting documentation. Regarding the outcome of the "LLM usage in Debian" option 5 "Responsible Use of Generative AI" [1], my understanding is that Debian contributors can use LLMs with their responsibility. However, I am not sure whether there is an agreed position specifically concerning their use for bug reports. Would it be appropriate to state this in the reporting guidelines for AI user? [1] https://www.debian.org/vote/2026/vote_002 On 19/08/2026 5:12 pm, Yunseong Kim wrote: > Hi Ben, > > On 19/08/2026 3:56 pm, Ben Hutchings wrote: >> This week's meeting will happen today (2026-08-19) at 21:00 CEST = 19:00 >> UTC. This time it will be on Jitsi if possible, otherwise we will fall >> back to IRC. >> >> The agenda is here: >> https://salsa.debian.org/kernel-team/meetings/-/wikis/20260819 >> >> I will attempt to add summaries of each bug's status before the meeting. >> >> Please reply to this message if you want to add items. > > Recently, I started drafting some guidance on avoiding unhelpful or > harmful AI-generated kernel bug reports, including reports submitted > without using reportbug. > > Do you think it would be worthwhile to add a section to the kernel > handbook [1], extend the DebianKernelReportingBugs wiki page [2], or > perhaps create a dedicated page covering this topic? > > A rough draft is below: > > --- > > = Debian Kernel Bug Reporting Etiquette: AI-Generated Reports = > > AI tools can be useful for improving writing quality, but they should > not replace the factual information needed for diagnosing kernel issues. > > Submitting unverified AI-generated output to the bug tracker or kernel > mailing lists is not a substitute for a proper bug report. Maintainers > and developers volunteer significant amounts of time to investigate > reported issues, and providing fabricated, guessed, or AI-invented > technical details wastes that effort. > > When reporting kernel bugs, please provide real observations, actual > logs, reproduction steps, and system information. > > Better yet, Please run: {{{ > $ reportbug kernel }}} > > The scripts in /usr/share/bug/linux-base/ and > /usr/share/bug/linux-image-*/ automatically collect information that > maintainers frequently need, including kernel taint status > (/proc/sys/kernel/tainted), loaded modules (lsmod), hardware details > (lspci -knn), and other diagnostic data [3]. > > If you are asking others to spend their time investigating a problem, > please provide accurate information gathered from the affected system. > Honest diagnostic data is far more valuable than unverifiable > AI-generated text. > > = References = > > [1] https://salsa.debian.org/kernel-team/kernel-handbook > [2] https://wiki.debian.org/DebianKernelReportingBugs > [3] https://kernel-team.pages.debian.net/kernel-handbook/ch-bugs.html#s9.1.1 > > --- > >> Ben. >> > > Best regards, > Yunseong Thank you for reading. I would appreciate any feedback. Best regards, Yunseong
