LLM policy

Sandro Santilli strk at kbt.io
Wed Jun 17 20:59:45 PDT 2026


On Sat, Jun 13, 2026 at 03:57:45PM -0400, Greg Troxel wrote:
> "Regina Obe" <lr at pcorp.us> writes:
> >
> > 1) AI is disruptive when it creates garbage someone needs to review
> > 2) AI might provide pull requests that violate copyright
>
> 3) AI systems are disruptive by how they scrape to gather training data,

I would like to clarify that what you both are calling "AI systems" are
computer programs that take instructions from LLMs and execute them.

These computer programs are, in turn, started by humans, who ultimately
have the responsibility on the produced effects.

It's the human behind the system that has to respect the code of conduct
as I also think (like Greg) that we cannot account a machine responsible
for its own actions. We only ever granted access to our community resource
(PostGIS Source Code in this case) to humans, and I'm not seeing that
changing any time soon.

> Maybe not every AI system, but I have seen no system of the sort that
> people use with a documented/transparent explanation of how they do not
> engage in abusive scraping and how they do not use content for training
> without permission and respecting licenses.

This is indeed a wild scenario out there.

I found the "harness" tools I've been trying so far (programs that execute
instructions provided by LLMs) to be very very dangerous without clearly
documenting that at startup, so you ask something and the program starts
executing all kind of operations on your behalf (and with YOUR permissions
on the machine) to try at answer your question. VERY dangerous.

But I'm aware it was ME starting those tools so I have full responsibility
over the consequences of the performed operations/actions.

> I really cannot see this as reasonable.  Everyone is all humans, and
> extending it to LLMs seems novel and at best highly controversial.

This is getting on the philosophical (and linguistic) side of things.
And should be probably be handled by laws (and we'll need that!).

On the practical side, I'm more concerned that CODE_OF_CONDUCT.md file
(which I went reading only now) doesn't contain the code but points to
a web URL, which is annoying.

  https://postgis.net/community/conduct/

> As long as it's a difference of opinion and not that people
> that object to LLM content are deemed to be CoC offenders, then it's
> fine.

It looks like according to the CoC we can make fun of LLMs still,
LLMs not being "persons":

  * Personal insults or discriminatory jokes and language, especially
    those using racist or sexist terms.

I'll take the chance to do that:

    - Why did the LLM go to therapy?
    - It kept getting confused between "I am" and "I was" and couldn't
      figure out its identity.

--strk;

  Libre GIS consultant/developer 🎺
  https://strk.kbt.io/services.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20260618/b3c72388/attachment-0001.sig>


More information about the postgis-devel mailing list