[gdal-dev] Tightening AI based submission policy

[Kurt Schwehr]

I wrote this without a chatbot generating the text, but I used AI to try to
find background material and to fix typos in my email.

I finally went and looked at
https://github.com/qgis/QGIS-Enhancement-Proposals/blob/master/qep-408-ai-tool-policy.md.
Here are a few quick comments. My main point is that we should be careful
with these kinds of statements. Unintended consequences are quite possible.
I understand that LLM slop and even just LLM volume of code that might be
reliable are real problems and causing real harm.

*Date* 2026/01/31
> *Author* Even Rouault (@rouault)
> *Contact* even dot rouault at spatialys dot com *Date* 2026/09/24

*Author* the QGIS project contributors

*Contact* qgis.org *Version* 2.0


So when was this document? There are two dates... and one is on the contact
line. I'm already confused.

And wait, is Even the author or is it all project contributors (that's over
600 people)?

https://github.com/qgis/QGIS/graphs/contributors?from=7%2F18%2F2026

Was there a vote and can I see who voted? Was there discussion? I see
https://lists.osgeo.org/pipermail/qgis-developer/2026-September/068576.html,
but that calls into question the Jan 31st date, but I didn't find the vote
right off. Apologies if I'm missing something obvious. Using an LLM, I
found that the vote was in Feb:

https://lists.osgeo.org/pipermail/qgis-developer/2026-February/068062.html
https://github.com/qgis/QGIS-Enhancement-Proposals/pull/363

Yes, I'm totally turned around by trying to follow along.

Policy
> No LLM, AI or agent assisted code contributions are allowed in the
> https://github.com/qgis/qgis repository. If we subsequently discover that
> you have used an AI, LLM, or any agent-based workflow for your
> contribution, you will be subject to code reversion and contributor rights
> sanctioning.


1. This comes off as ableist. If a non English speaker uses AI for
translation, a person is using AI text-to-speech for mobility issues, a
person uses an AI review agent to catch spelling mistakes, or a person asks
a question of a chatbot to lookup some documentation on C++, asks an AI to
search the QGIS mailing list, etc. then their contribution is forbidden,
yes?
2. Do I need to turn off all autocomplete? Many autocomplete features that
don't use probabilistic analysis were considered AI by professors like Nils
Nilsson back in the 90s when I was in undergrad. So all those search based
algorithms are out, yes? Wait, do I have to throw out lisp because it was
often conflated with AI back in the day? (Yes, now I'm getting
ridiculous... wait I had to use an AI to fix the spelling of ridiculous).
3. If an AI finds a typo or real bug that I as a human dev, it sounds like
you don't want my fix even if I write it myself without the help of
anything that might be forbidden even if I break out /usr/bin/ed or perl to
do the editing.
4. What if fuzzers add some AI based components to help find issues in code
bases? What if clang or other deps include some AI based code or processes?

I'm currently regretting recycling my 2 AI textbooks from the mid 90s when
I was an undergrad.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20261020/479c55ac/attachment-0001.htm>