Re: [PATCH] gitk: discourage AI contributions

[Weijie Yuan <wy@wyuan.org>]

[+cc Johannes Schindelin]

On Sun, Aug 23, 2026 at 11:25:42AM +0200, Johannes Sixt wrote:
> Since the provenance of code generated by AI cannot be tracked with
> confidence, which would be necessary to sign off a contribution, tell
> contributors not to use AI.
> 
> Signed-off-by: Johannes Sixt <j6t@kdbg.org>
> ---
>  Following a recent brief discussion[1] it looks like I should not
>  accept code for Gitk that was generated by AI even if the submitter
>  understands all of it. Should I discourage AI contributions outright
>  like this or would it overshoot the mark?
> 
>  -- Hannes
> 
>  [1] https://lore.kernel.org/git/aooRdiVdjovWSFiG@fruit.crustytoothpaste.net/T/#md44580584f30c9b3e3c9e23250baf08b7fd7fb0c

I followed the discussion between Brian and Junio, and I have no
intention of further expanding the discussion about the legal and
technical issues surrounding the use of AI/LLM. However, I would like to
point out that several of dscho's recent commits seem to carry an
Assisted-by trailer, and, unless I missed something (possibly), nobody
appears to have raised the issue or strongly objected to it. Of course,
I believe dscho also fully understood the code he submitted.

(This is getting a little away from the DCO legal issue that Junio and
Brian were discussing.)

So how do we distinguish between "Generated" and "Assisted"? Perhaps the
LLM merely assisted him with some investigative work rather than
actually writing any code.

Meanwhile, in the Linux kernel community, probably the most prominent
user of the DCO, within it, Linus also expressed his views [1] on the
use of LLMs a few weeks ago, while the Linux Foundation has published
guidance on the legal issues surrounding the use of AI. [2]

Although I know that the Git and Linux kernel communities do not follow
exactly the same rules, Jonathan Corbet updated the kernel's
requirements for disclosing LLM use a couple of days ago, namely:

Assisted-by: LLM [TOOL1] [TOOL2]

Perhaps this gives us a way to tell whether a patch contains content
generated by an LLM. (But if we want to ban it entirely, please
disregard this part.)

Thanks,
Weijie

Emm, I´m not sure which thread would be the right place to continue this
discussion, sorry.

[1] https://lore.kernel.org/all/CAHk-=wi4zC+Ze8e+p3tMv8TtG_80KzsZ1syL9anBtmEh5Z40vg@mail.gmail.com/
[2] https://www.linuxfoundation.org/legal/generative-ai