Re: Super Simple AI Policy proposal

Akseli Lahtinen <[email protected]> Mon, 24 Nov 2025 11:56:00 +0200
Newsgroups	gmane.comp.kde.devel.core
Message-ID	<5025580.GXAFRqVoOG@desktop>
On Monday 24 November 2025 11:25:00 Eastern European Standard Time Vlad 
Zahorodnii wrote:
> On 11/24/25 11:20 AM, Akseli Lahtinen wrote:
> > On Monday 24 November 2025 11:09:19 Eastern European Standard Time Vlad
> > 
> > Zahorodnii wrote:
> >> On 11/24/25 2:16 AM, Akseli Lahtinen wrote:
> >>> Contributor is always responsible for the code changes and creation,
> >>> regardless where the code came from, such as:
> >>> - Code was written completely by the contributor.
> >>> - Code was generated by an LLM/"AI" or any other tool.
> >>> - Code was given to contributor by someone else.
> >> 
> >> It needs to be supplemented with a license.
> >> 
> >>> - Code was copy-pasted from the internet.
> >> 
> >> If the code has a license attached to it, sure. Otherwise, it seems like
> >> a no-no thing. That being said, you can't also verify that that code
> >> hasn't been copied from elsewhere, but I don't think that we should say
> >> it's okay to do it. You can have a look at code and have your own take
> >> on it.
> >> 
> >>> Contributor must try their best to understand what their contribution
> >>> is changing and they must be able to justify the changes.
> >>> 
> >>> Using any tools to help understand and justify those changes does not
> >>> change or reduce the expectations.
> >>> 
> >>> The changes are attributed to the contributor, no matter whatever tools
> >>> they have used.
> >> 
> >> I am not a lawyer and I don't follow the legal stuff closely, but last
> >> time I heard about the ownership attribution cases, it had still been a
> >> muddy water thing. There are several schools of thought, each of which
> >> has a leg to stand on: "it's a clever autocomplete machine, so you can't
> >> claim ownership of a work, which is owned by somebody else" or "you
> >> can't claim copyright just because you typed a query, you didn't put in
> >> enough of creative effort to claim the ownership," or "the generated is
> >> not strictly the same as the source data so it has been transformed,"
> >> etc.
> >> 
> >> To be frank, I think the best course of action is to sit it out and see
> >> where laws end up being.
> > 
> > I am definitely fine with this too. My main concern is, what do we say to
> > people asking about this?
> 
> As you said, if a person doesn't understand the changes at all, I think
> such MRs should be rejected. But figuring it out will be a cat and a
> mouse game (because they could use the LLM to answer your questions :p).
> That being said, I personally don't think KDE should come out and say
> "we approve/reject AI", it's still uncharted territory, maybe some vague
> guidelines for maintainers and developers but nothing strict.
> 

Hm, so instead of direct policy, just have maintainer guideline for dealing 
with this situation?

I do like that idea actually!

> Regards,
> Vlad

Best regards,
- Akseli
signature.asc (application/pgp-signature, 659 B)