On 25/07/26 at 16:09 +0100, Ian Jackson wrote:
> Andrea Pappacoda writes ("Re: GR: Ban LLM contributions from Debian"):
> > I want to add, though, that I'm not sure how I feel about preventing LLM
> > usage for translation.
>
> As well as all the terrible ethical problems with LLMs, modern
> LLM-generated "translations" have the same problem that LLM
> "summaries" etc. have: they just make it up if they don't know.
>
> I don't know if any previous-generation more-reliable[1] machine
> translation is still available.
>
> > Also, point 8 is in direct conflict with Planet's
> > rules and the mailing lists code of conduct, which mandate English usage.
>
> With this proposal I don't intend to overrule the Planet admins or the
> listadmins. I am (perhaps pusilanimously) trying to write a proposal
> which preserves the autonomy of maintainers, so that they'll be
> allowed to burn the world to hell if they want to.
>
> How about:
>
> 4. Within Debian, messages to humans (including for example bug
> reports, mailing list messages, discussions on Salsa, and blog
> posts on Planet Debian) must be drafted solely by humans without
> LLM assistance. However, policies published by relevant package
> maintainers or communication forum administrators shall take
> precedence.
I recently produced two concrete cases that might be worth considering
to help define what (4.) should say.
First, in https://lists.debian.org/debian-vote/2026/07/msg00017.html, I
provided pointers to AI-generated summaries of previous discussions on
this topic. The mail is fully written by me, but it advertises
AI-generated content (that is clearly labelled as such in the mail).
Do you think that this should be evaluated as a violation of the CoC?
Second, in #1142332, I reported a bug about lintian hanging on some
package. I could have stopped there, but instead asked an LLM to analyze
the issue, and to produce a bug report. The part of the bug report above
"====" is written by me without AI assistance. the part below is written
with AI assistance.
Do you think that this should be evaluated as a violation of the CoC?
If I only reported the AI findings (writing them down myself, without
including the AI-generated bug report), should it still be a violation
of the Coc?
If I reported the AI findings and pointed to the AI-generated bug report
hosted elsewhere (without including it directly in the bug mail), should
it still be a violation of the Coc?
Lucas
