On Mon, Aug 10, 2026 at 11:41:41AM -0700, Soren Stoutner wrote: > On Monday, August 10, 2026 9:47:53 AM Mountain Standard Time Bdale Garbee > wrote: > > Andrey Rakhmatullin <[email protected]> writes: > > > On Mon, Aug 10, 2026 at 10:48:44AM +0200, Matthias Urlichs wrote: > > >> the "bah so > > >> > > >>many choices why do I even bother" effect is still a concern > > >> > > > At this point I'm afraid it's inevitable for most GRs and we can't do > > > anything for that short of changing the culture (and I think there is > > > always someone who says "we should have many choices" in this GR) or > > > providing summaries and comparisons. > > > > To me, a GR with many options is evidence that our collective > > understanding of a topic has not yet reached the point where a GR > > is the obviously correct next step. > > That’s the beauty of the Further Discussion option. If there isn’t enough > consensus in Debian for any of the options to be selected, then Further > Discussion wins. That, in and of itself, is both a valid and useful outcome. > > It says that the project has selected the status quo above any of the other > options. > > Among other things, if the project selects Further Discussion, it would mean > that no one should feel justified in attacking any other member of the > project > for using AI as long as the output quality met the project’s standards and > any > generated code is DFSG-free.
But when a maintainer chooses to ignore contributions with AI assistance, we will have the AI discussion at the Technical Committee. Vulnerability reports without any AI assistance are becoming rare, and based on this GR discussion I have become worried that maintainers of a trusted piece of our infrastructure might refuse to even look at vulnerabilities found/reported with AI assistance. > Because options that would allow banning such AI > use would have lost to Further Discussion. No, rejection of a ballot option does not make any decision. Like for me the issue below is sufficient reason to make option 5 worse than the default option, but this does not imply I would be opposed to "Responsible Use of Generative AI". > What we would end up with is the status quo, which is that every Debian > Developer can choose how they want to engage with AI. In many regards, > option > 5 is an attempt to capture the status quo in writing. The difference between > option 5 winning and Further Discussion winning is that option 5 would make > it > more explicit that the project as a whole is generally satisfied with the > status quo, making it less likely to change in the future (although option 5 > explicitly states that the project’s views towards AI may change in the > future, including without the need for GR). Because we have an option that > simply codifies the status quo, if Further Discussion wins that means that > the > project is not satisfied with the status quo and expects it to change in the > future, but is uncertain as to how it should change. >... Option 5 is not the status quo. Option 5 creates a right for contributors to not disclose AI usage: We enourage our contributors to disclose whether a contribution was made with AI assitance, but do not require them to do so. If this is what Debian decides, then maintainers must not force disclosure by asking contributors about AI assistance. cu Adrian BTW: There are 2 misspellings in the GR sentence I quoted. Proofreading of GR texts might be a good idea.
