Re: GCC GSoC 2026: Call for project ideas and mentors

On Mon, 19 Jan 2026 at 21:00, Toon Moene <[email protected]> wrote:
>
> On 1/19/26 19:51, Richard Biener wrote:
>
> > On Mon, Jan 19, 2026 at 7:46 PM Ville Voutilainen
> > <[email protected]> wrote:
>
> > The DCO, even though I think it's obvious, might also get a clarifying
> > amendment that what an AI tool produced is not obviously your work
> > or can be considered OK to contribute.  Same holds for people
> > contributing using FSF assignment - you can only assign sth you
> > have rights on.
>
> This might seem far fetched for people who do not know how Machine
> Learning works - but it is not impossible that the suggested code it
> comes up with is in whole or in a very large part someone's actually
> published work.
>
> My own brother encountered this when quotes from his text book on
> air-surface interaction in boundary layer meteorology showed up in
> students' work *that could be traced back* to the use of AI tools.

There's also more and more newbies who don't grok that, and work based
on hearsay. "I thought LLM output
isn't considered copyrightable", I heard today from a very surprising
source. I did explain that it's not about
someone slapping some additional copyright on a tool's output, but the
tool regurgitating copyrighted material
it has used as some sort of input. That explanation rang all the right
bells quickly enough, but going forward,
I think it wouldn't hurt to remind people about what's what more often
and in more places.

I also heard today from a slightly surprising source that there are
places where the no-LLM policies say "doesn't
apply to documentation" and I need to go and fix their misconceptions. :D